Has Whole Language Failed?

Stephen Krashen

University of Southern California

Rossier School of Education

January, 2000

(Reprinted with permission of the author)

The "failure" of whole language in California has been widely reported. I attempt here to give a clear definition of whole language, discuss some of the research, and provide some information about the impact of whole language in California.

What is Whole Language?

There are several competing definitions. One use of the term "whole language" refers to what we will call the Comprehension Hypothesis. Other definitions are very different, even contradictory to the Comprehension Hypothesis.

The Comprehension Hypothesis

The Comprehension Hypothesis (a.k.a. the Input Hypothesis, Krashen, 1985) claims that the development of literacy and the development of language in general occur in only one way: When we understand messages. Smith (1975) stated the Comprehension Hypothesis in the title of his book: Comprehension and Learning, claiming that comprehension and learning are very much the same thing. Reading pedagogy, according to the Comprehension Hypothesis, focuses on providing students with interesting, comprehensible texts, and the job of the teacher is to help children read these texts, that is, help make them comprehensible. The direct teaching of "skills" is helpful only when it makes texts more comprehensible.

More precisely, comprehension of messages is necessary for language acquisition and literacy development, but it is not sufficient. It is certainly possible to comprehend a text or message and not acquire anything. We acquire when we understand messages that contain aspects of language that we have not yet acquired but are developmentally ready to acquire.

The Comprehension Hypothesis claims that we learn to read by reading (Goodman, 1982; Smith, 1994a), and that other aspects of literacy competence are the result of meaningful reading. Reading, it is claimed, is the source of much of our vocabulary knowledge, writing style, advanced grammatical competence, and spelling.

Whole language has been defined in several other ways:

The Incomprehensible Input Hypothesis

Incredibly, whole language has been described as just the opposite of the Comprehension Hypothesis: Providing incomprehensible input. Los Angeles Times reporter Richard Colvin (1995) characterizes whole language as giving children texts they do not understand:

"The frustration of students taught with the whole language method was obvious last year in the faces of her first graders, said Tammy Hunter-Weathers, a teacher at Hyde Park School in the Crenshaw area of Los Angeles. 'The children were in tears,' she said, when they were asked to read texts even though they did not know the letters or sounds. 'They look at you with three paragraphs on a page and they say, 'What do we do with this?''"

This is not the Comprehension Hypothesis version of whole language. Practice based on the Comprehension Hypothesis will focus on providing interesting and comprehensible texts, and an important role of the teacher is to help children understand them.

The Output Hypothesis

Delpit (1986) has argued that "holistic teaching approaches" do not give minority children competence in the forms "demanded by the mainstream' (p. 383), and recommends the teaching of skills in context. Delpit's view of whole language, however, was instruction that emphasized a great deal of writing: "I focused energy on 'fluency' and not on 'correctness'" (p. 381).

Delpit's conclusion that a focus on writing fluency will not do the job, although not supported with data or writing samples in her report, is consistent with research on writing (Krashen, 1993). There is no support for the hypothesis that writing in of itself causes language acquisition or literacy development. Acquisition of the conventions of writing, it has been argued, is a result of reading, not writing (Smith, 1994b; Krashen, 1993). Delpit does not mention whether reading was emphasized in her classes.

The Comprehension Hypothesis does not suggest that we avoid writing. There is evidence that writing, while not a means of language development, is a powerful way of clarifying thinking (Krashen, 1993).

Look-Say

Shanahan, the CEO of Gateway, the company that produces Hooked on Phonics, claimed that his son had been taught "with what was known as the whole language method and was expected to remember hundreds of whole words by their shapes, with occasional clues from pictures or context ..." (Shanahan, 1994, p. 3). This is neither whole language nor an application of the Comprehension Hypothesis, but is Look-Say, a method which focuses on the memorization of sight words.

No Phonics?

The Comprehension Hypothesis does not forbid the direct instruction of phonics. Weaver (1994) and Krashen (1996) have pointed out that proponents of phonics typically support the teaching of just the straight-forward phonics rules, and expect children to "induce" the more complex rules. This is exactly the position of those sometimes considered to be anti-phonics. There is surprising agreement when one looks at the research. Smith's conclusion (Smith, 1994a) appears to be the most reasonable: Teach "skills" when they help make texts comprehensible. It is, of course, an empirical question just how useful direct teaching of phonics is in making texts comprehensible.

What is Whole Language?

The term "whole language" does not refer only to providing interesting comprehensible texts and helping children understand less comprehensible texts. It involves instilling a love of literature, problem-solving and critical thinking, collaboration, authenticity, personalized learning, and much more (Goodman, Bird, and Goodman, 1991). In terms of the process of literacy development, however, the Comprehension Hypothesis is a central part of whole language.

Does Whole Language Work?

The claim has been made that skills-based approaches produce results superior to whole language approaches. The origin of this claim is Chall (1967), who concluded that methods that stressed systematic phonics instruction were superior to methods that stressed intrinsic phonics (less phonics, and in context), and that both systematic and intrinsic phonics were superior to Look-Say, which involved no phonics at all. None of these comparisons dealt with the kind of whole language considered here, that is, methods that emphasize a great deal of interesting, meaningful reading.

After a review of more recent studies of method comparisons involving beginning readers, I have concluded that when "whole language" is in fact real reading, it does very well. Students in classes that do more real reading have better attitudes toward reading (McKenna, Stratton, Grinkler, and Jenkins, 1995; Merver and Hiebert, 1989), read more (Freppon, 1995), do as well as traditional students on tests in which the focus is on form, do as well or better on more communicative tests (Merver and Hiebert, 1989; Hagerty, Hiebert and Owens, 1989; Morrow, O'Connor and Smith, 1990; Klesius, Griffith, and Zielonka, 1991; Morrow, 1992) and show better development of the kind of language used in books (Freppon, 1995; Purcell Gates, McIntyre, and Freppon, 1995). Foorman, Francis, Beeler, Winikates, and Fletcher (1997) is the only apparent counter example to the generalization that students who do more real reading will outperform those who do less. It was not clear, however, that the "whole language" children did more reading than the other children; the abysmal scores in reading comprehension for all subjects in this study suggest that none of the groups did much reading.

On the other hand, when whole language is not defined as real reading, it does not do well when compared to skills-based methods. Here are some examples: - In Holland and Hall (1989), no differences were found between whole language and a basal method, but the whole language method emphasized deliberate vocabulary development and a focus on words in isolation. - In Reutzel and Cooper (1990), results favored whole language, but both groups read a great deal. - In Eldridge and Baird (1996), a "phonemic awareness" approach was claimed to be superior to whole language, but "whole language" included "studying" words and sentences in stories, and the children in whole language "were taught to read" using a basal reader (p. 198).

Stahl, McKenna, and Pagnucco (1994) reported in their meta-analysis that whole language students were better in four studies, traditional methodology was better in one study, and no difference was found in 12 studies. While all studies analyzed by Stahl et. al. are listed in their bibliography, they do not tell us which studies were used in their analysis. In addition, many of the studies are unpublished. We thus have no idea what "whole language" meant in this analysis.

Method comparison studies thus show that whole language is not a failure: On standardized tests, children who do more real reading do as well asor better than children who read less, and consistently do better on other measures, a result very similar to that found for sustained silent reading for more advanced readers (Krashen, 1993).

A Decline in California?

There is the perception that reading has gotten worse in California, because California fourth graders did so poorly on the recent NAEP test, compared to other states, and in 1987 a literature-based approach was officially endorsed by the state. But there is no evidence that reading scores have declined in California, as shown by CAP scores from 1984 to 1990:

from: Guthrie et. al., 1993, cited in McQuillan, 1998.