Brown Et Aldoc

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

A PRELIMINARY STUDY OF CLOZE PROCEDURE AS A TOOL FOR

ESTIMATING ENGLISH READABILITY FOR RUSSIAN STUDENTS

JAMES DEAN BROWN, GERRIET JANSSEN, JONATHAN TRACE


University of Hawaii at Mnoa
&
LIUDMILA KOZHEVNIKOVA
Samara State University
ABSTRACT
This preliminary study examines the relationships between each of six first language (L1) readability indexes
and the cloze passage mean performances of Russian EFL students. The cloze passages were created by
randomly selecting 50 text passages from an American public library and deleting every 12th word in each
passage to create a 30-item cloze procedure. The participants were 5170 EFL students from 38 universities in
the Russian Federation. Each student was randomly assigned to take one of the 30-item cloze passages. The L1
readability indexes calculated for each of the 50 passages were the Flesch, Flesch-Kincaid, Fry, Gunning, Fog,
and modified Gunning-Fog indexes. The preliminary results indicate that the L1 readability indexes were
moderately to highly correlated with each other, but only somewhat correlated with the mean performances of
Russian university students on cloze versions of those same passages. These results are discussed in terms of
why the L1 readability indexes are moderately to highly correlated with each other but only somewhat
correlated to the Russian EFL means. The authors also explain what they are planning in terms of further
linguistic analyses (e.g., of variables like average word length, percent of function words, number of syllables
per sentence, number of words per paragraph, frequencies of words in the passages, and so forth) and statistical
analyses (including at least factor analysis, multiple regression analyses, and structural equation modeling) of
these data.

INTRODUCTION

First Language Readability


Readability is a concept that describes the degree to which a text is easy or difficult to read.
A readability index is a numerical scale that estimates the readability or degree reading difficulty
Second Language Studies, 31(1), Fall 2012, pp. 1-22.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

that native speakers are likely to have in reading a particular text. For example, the Fry (1977)
readability scale is expressed in grade levels for students in the United States ranging from 1
(first grade) to 17+ (graduate school and beyond). Thus a passage with a Fry scale index of 3.5,
would be fairly easy because it would be appropriate for children who are native speakers of
English in the second half of third grade (or about 8 years old), whereas a passage with an index
of 13 would be more difficult because it would be suitable for first-year university-level native
speakers of English.
The findings from one study (Brown, Chen, & Wang, 1984) led the first author believe that
such L1 readability indexes might be useful indicators of relative passage difficulty in EFL
settings. Brown, Chen, and Wang studied the readability of the cards in Stanford Research
Associates (SRA) classroom reading kits. Those kits have cards at different grade levels (coded
by color) that had previously been established by research into the actual performances on those
cards of North American children. Brown, Chen, and Wang started by calculating the Fry
readability index for each of the SRA cards. They then compared the resulting Fry scale indexes
with the actual native-speaker grade-level performances.
Table 1
Accuracy of L1 Fry Readability Estimates
(Adapted from Brown, Wang, & Chen, 1984)

SRA
Kit
3A

4A

Passage
Grades Based
on Student
Performance
3.5
4
4.5
5
6
7
8
9
10
11
8
9
10
11
12
13
14

Mean
3.22
4.56
5.56
6.44
7.11
8.22
8.67
9.56
10.22
10.11
8.56
9.44
10.44
11.11
12.56
13.11
13.25

SD
1.20
1.42
0.88
0.73
0.93
2.17
1.50
1.67
1.48
2.15
1.13
0.88
1.74
1.83
1.51
3.30
1.98

Min
2
3
4
5
6
6
6
6
7
6
6
8
9
7
11
9
9

Max
6
6
7
7
8
13
10
12
12
12
10
10
14
13
16
17+
15

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

Table 1 shows results for the 3A and 4A SRA kits separately. The grade levels shown in
column two were established by research into the performances of L1 native speakers (each
grade level consisted of 12 to 14 cards). The mean1, standard deviation2 (SD), minimum3 (Min),
and maximum4 (Max) for the Fry scale readability estimates for each grade or half-grade level
are shown in the last four columns. Notice, in the third column, that the mean grade levels for the
Fry index are remarkably close to the actual grade levels of the cards as established by student
performance. Clearly, this study demonstrates that there is a strong relationship between the
mean grade levels estimated by the Fry scale and the grade levels established on the basis of
native-speaker performances.
A large number of English as a first language (L1) readability indexes have been invented
over the past 60 years. Chall (1958), Klare (1963; 1984), Zakaluk and Samuels (1988), or
Zamanian and Heydari (2012) all provide overviews of the first language readability literature.
The L1 readability indexes examined in the current paper are: the Flesch reading ease formula
(Flesch, 1948), the Flesch-Kincaid readability index (as described in Klare, 1984), the Fry
readability scale (see Fry, 1985), as well as the Gunning, Fog, and modified Gunning-Fog
readability indexes (see Larson, 1987).
The simplest way to explain the L1 readability indexes is to show the equations that define
them. For example, Fleschs (1948) equation multiplies the average number of syllables per
word in the text by .846, then subtracts the result from 206.835. From this result, the equation
subtracts 1.015 times the average number of words per sentence. The actual equation for the
Flesch reading ease index is:

1. Flesch Reading Ease Formula (Flesch, 1948)


= 206.835 - .846 (syllables / words) - 1.015(words / sentences)

Here the mean can be interpreted as the more familiar arithmetic average.
The standard deviation is a sort of average of the distances from the mean of all the values in the data; as such, it is
an indicator of how much the values are dispersed around the mean.
3
The minimum is the lowest value in the set of numbers.
4
The maximum is the highest value in the set of numbers.
2

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

The other indexes described below (numbers 2 to 6) are similar manipulations of the numbers of
syllables, words, long words, easy words, hard words, sentences, etc. For those who are
interested, we also provide a seminal reference for each equation.

2. Flesch-Kincaid Index (as cited in Klare, 1984)


= .39(words / sentences) + 11.8(syllables / words) - 15.59

3. Fry Scale (Fry, 1977 or 1985)


= on the Fry reading graph (see Fry, 1985), the grade value at the point where the
coordinates for sentences per 100 words and syllables per 100 words cross

4. Gunning Index (as cited in Carrell, 1987)


= .40(words / sentences + % of words over two syllables)

5. Fog Count (as cited in Carrell, 1987)


+ 3( )
3}

=
2
{

6. Gunning-Fog Index (Larson, 1987)


={

/
+ 100} {
}

Second Language Readability


In contrast to the rather large literature on L1 readability indexes, very little work has been
done on readability indexes applied to second language (L2) students. A few such studies have
investigated readability in languages other than English. For example, Nguyen and Henkin
(1982) did so for Vietnamese, and Gilliam, Pea, and Mountain (1980) did so for Spanish.
Moreover, Klare (1963) provided a survey of nine other early readability studies for French,
German, Japanese, and Spanish.
For ESL, Haskell (1973) found that cloze procedure successfully distinguished among texts
regardless of their length, the scoring method used, the deletion rate, and so forth. Hamsik (1984)
found fairly strong associations (ranging from .78 to .82) between student performances on cloze

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

tests and four different readability indexes. Brown (1998) showed that the mean performances of
Japanese university students on the same 50 passages used in the present study correlated with
the same readability estimates used in this study ranging from .48 to .55. Greenfield (1999)
replicated Brown (1998) with different passages and found that the traditional L1 readability
indexes correlated strongly with Japanese students performances on cloze tests. Greenfield
(2004) reported similar results and concluded that the traditional L1 readability indexes are
valid for EFL use (p. 5).

Cloze Procedure and Readability


The first reference to cloze procedure was Taylor (1953), who studied the value of this sort
of test for estimating the readability of reading materials used in U.S. public schools. Over the
ensuing years, other key studies on cloze readability have included Bickley, Ellington, and
Bickley (1970), Bormuth (1966, 1968), Brown (1998), Greenfield (1999, 2004), Miller and
Coleman (1967), Moyle (1970), Ransom (1968), and Taylor (1957). All of these studies have
shown that performances on cloze tests are at least somewhat related to readability.

Purpose
The primary purpose of this study was to explore the degrees of relationship between each of
the L1 readability estimates and Russian EFL students performances on actual cloze passages.
To that end, the following research questions were posed:
1. Are randomly selected cloze tests reliable and valid tools for gathering data on passage
difficulty?
2. To what degree are traditional first language readability indexes related to the average
cloze scores for the same passages (when administered to Russian EFL students)?

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

METHOD

Participants
This study focused on the performances of 5170 Russian EFL students. The participants
were selected as intact EFL classes from 38 different universities across Russia.5 The
participants ranged in age from 14 to 45 with a mean of 18.59 (48 participants did not answer
this question); 71.7% of the participants were female, 28.0% males, and 0.3% did not specify
their gender. All 50 cloze procedures were administered in such a way that all students were
randomly assigned across all testing sessions to particular cloze tests. The purpose of doing
this was to insure that the results of the different groups could reasonably be assumed to be
comparable across the 50 passages. An average of 103.4 students took each cloze test (with
individual passages ranging from 90 to 122 students).
The participants were mostly undergraduate students from non-linguistic universities and
departments, though, some students were linguistics majors. Their levels of proficiency can
generally be said to represent the English proficiency of university students in Russia who are
studying subjects other than English or linguistics. While the participants in this study were not
randomly sampled from all Russian university students, it can be said that the sample is fairly
large and homogeneous with regard to the nationality, language background, and educational
level of the students.

Materials
The 50 cloze procedures used in this study were developed by randomly selecting 50 books
from the adult reading section of the Leon County Public Library in Tallahassee, Florida, and
then randomly choosing a passage from each book to create a 350 to 450 word long passage,
beginning from a semantically logical starting point. Clearly, these passages were not selected in
a completely arbitrary manner, but they were selected so that they would form independent and
cohesive passages. The resulting 50 cloze passages ended up ranging from 366 to 478 words

We would like to thank all of our colleagues who helped at various stages of this project by administering the cloze
tests at 38 universities in the following 25 towns and cities: Chelyabinsk, Kazan, Kolomna, Krasnodar, Moscow,
Novocherkassk, Novosibirsk, Orenburg, Rostov/Don, Ryazan, Samara, Saransk, Saratov, Smolensk, Solykamsk, St.
Petersburg, Surgut, Syktyvkar, Syzran, Taganrog, Togliatti, Tomsk, Ulyanovsk, Voronezh, and Yoshkar-Ola. For a
list of the cooperating institutions, see Appendix A.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

with an average of 412.1 words in each passage. Based on random selection, the resulting set of
50 passages is assumed to represent the reading passages encountered in U.S. public library
books.
To create the cloze passages, every 12th word was deleted from each text and was replaced
with a standard length blank. A 12th word deletion pattern was chosen instead the more
traditional 7th word deletion so that 30 items could be constructed far enough apart to minimize
the effect of answering one item correctly (or incorrectly) on answering other items. In addition,
one sentence was left unmodified by blanks at the beginning of each passage and one or two
were left intact at the end of each passage. Additional spaces were then added at the top for the
students name, sex, age, native language, and country of passport. Directions were also given
that explained what the students must do to fill in the blanks and how they would be scored.
Sample directions and 12 sample items are shone in Appendix B, which was taken from Brown
(1989).
An additional very short 10-item cloze passage was also created and attached to all 50 of the
cloze tests. This 10-item cloze test was developed on the basis of pretesting reported in Brown
(1989), using procedures similar to those applied in Brown (1988), so that only those blanks that
had proven very effective in an item analysis were kept in the test. The purpose of this short
anchor test was to provide a common metric for making comparisons among the fifty groups of
students and for anchoring item response theory analyses, which will be reported in future
studies.

Procedures
The data for this paper were collected by a large number of teachers at 38 universities in
various locations throughout the Russian Federation (see footnote 5 for those locations). The
cloze passages were randomly distributed in a manner that assured that all students had an equal
chance of getting any of the 50 cloze tests. They were administered by the teachers in classroom.
The directions were clarified as necessary, and a total of 25 minutes was allowed for completing
both the 30-item and ten-item cloze tests. According to teacher feedback, 25 minutes was
sufficient time for students to finish.
Exact-answer scoring was used in this research. Exact-answer scoring involves counting only
the original word that had occupied the blank as a correct answer. We felt that this was justified

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

because research has repeatedly shown high correlations between exact-answer scores and other
more elaborate scoring procedures (Alderson, 1978, 1979; Brown, 1978, 1980).

Analyses
The variables in this study were chosen because, they were known to be related to readability
and because they were quantifiable. In other words, these variables were chosen because they
might explain statistically the variations in readability levels of the cloze passages in this
research. Only seven variables are included in this preliminary report: (a) six L1 readability
indexes and (b) the means produced by the Russian EFL students who took these tests.

PRELIMINARY RESULTS

Table 2 shows the descriptive statistics for the 50 cloze tests (k = 30). These statistics include
the mean, standard deviation (SD), minimum score obtained (Min), maximum score (Max), the
number of participants who took the particular cloze (N), and the Cronbach alpha internal
consistency reliability estimate for each test.
One obvious result which jumps out of Table 2 is that the means for the 50 cloze passages
range from a low of 1.60 to a high of 12.82 out of 30. Since random selection promotes the
equality of these groups in terms of overall English proficiency, the variation in means revealed
in Table 2 is probably due to considerable variation in the difficulty levels of the passages
involved. It is worth noting that these means are fairly low for tests with 30 items each.
However, similarly low means have been commonly reported for cloze tests scored using the
exact-answer method.
Notice also that the standard deviations range from a low of 2.27 to a high of 6.96. This
range indicates substantial variation in the degree to which the students scores were spread out
around the means of these cloze passages. The minimum (Min) values are all 0. The maximum
(Max) values ranged from 10 to 30, which indicates substantial variations in the ways these cloze
passage scores spread out around their respective means. The number of participants on each
cloze passage also ranged from 90 to 122.6 The reliability of the 50 cloze tests likewise varied
6

Note that, for reasons that we do not yet understand, Passage 1 originally had an unusually high number of
participants with 170. We randomly selected 120 to remain in our data so that the sample sizes would all be
approximately the same.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

considerably. Notice that the lowest internal consistency Cronbach alpha reliability was .646,
while the highest was .919. Such reliability estimates indicate the proportion of reliability or
consistency in the scores. For example, .646 indicates that 64.6% of the variance in scores for
that cloze test was reliable but also, by extension, that 35.4% was unreliable (100% - 64.6% =
35.4%). Reliability estimates are important in any statistical research because a study can only be
as reliable as the measures upon which it is based.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY


Table 2
Descriptive Statistics for 50 Cloze Passages
Test
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

Mean
6.78
7.06
3.94
9.82
6.54
5.34
8.07
3.13
4.08
3.77
5.74
9.27
3.30
5.10
8.10
2.30
2.55
1.60
6.15
5.41
10.32
3.74
3.58
2.13
4.63
4.35
3.48
4.01
3.39
12.82
4.88
4.96
2.82
7.11
6.72
4.81
8.38
2.42
3.62
3.87
4.53
4.78
2.09
4.80
9.24
3.69
3.19
2.98
4.39
3.57

SD
3.99
4.94
3.71
6.12
4.38
4.19
6.22
3.67
3.67
4.24
4.53
4.86
3.89
4.70
5.60
2.70
2.29
2.27
5.08
5.01
6.96
3.64
3.36
2.37
4.55
3.25
3.07
3.81
2.70
5.39
3.89
3.22
2.57
4.43
5.54
4.11
5.46
2.44
3.44
4.39
3.56
4.10
2.56
4.28
6.59
3.49
2.79
3.36
4.10
3.04

Min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Max
16
19
14
21
22
16
20
24
23
22
17
20
17
17
21
11
10
15
30
24
27
14
14
10
15
21
15
18
11
22
14
12
10
18
25
16
24
14
12
24
14
20
15
19
21
14
12
18
15
13

N
120
102
103
105
106
102
103
101
105
102
101
115
105
107
106
115
109
100
102
97
103
102
102
101
102
100
100
102
102
111
101
101
102
102
103
96
103
106
103
90
101
93
99
102
101
93
104
108
122
109

Reliability
0.745
0.853
0.811
0.887
0.822
0.828
0.896
0.859
0.808
0.869
0.845
0.834
0.855
0.866
0.893
0.773
0.646
0.775
0.882
0.887
0.919
0.825
0.789
0.712
0.873
0.770
0.751
0.837
0.702
0.834
0.815
0.785
0.713
0.828
0.873
0.834
0.872
0.743
0.804
0.877
0.794
0.836
0.760
0.854
0.909
0.803
0.729
0.753
0.858
0.758

10

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

11

Table 3 displays the results for the L1 readability indexes examined in this research. They
are arranged not by the passage numbers as they were in the previous table, but rather in order
from the easiest to most difficult as indicated by the means in the last column. In other words,
high means on passages (like the 12.82 for passage 30) indicate that the Russian students found
them to be relatively easy, and low means (like the 1.60 for passage 18) indicate that the students
found them to be relatively difficult.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY


Table 3
L1 Readability Estimates and Russian Means for 50 Passages

Passage
30
21
4
12
45
37
15
7
34
2
1
35
5
19
11
20
6
14
32
31
36
44
42
25
41
49
26
9
28
3
40
10
22
46
39
23
50
27
29
13
47
8
48
33
17
38
16
24
43
18

Flesch
4.63
4.74
5.95
8.59
8.47
6.03
9.69
9.37
10.69
10.71
6.78
3.69
11.00
8.27
2.71
8.30
5.18
4.79
7.80
8.13
7.88
11.6
7.10
7.72
12.26
7.59
13.95
12.30
12.00
2.83
5.69
11.86
8.97
8.78
5.09
11.45
18.51
9.36
13.58
10.65
9.99
8.46
8.51
13.82
15.60
11.01
8.90
10.69
11.51
9.69

FleschKincaid
6.5
7.5
7.6
11.0
11.1
8.6
12.0
9.9
12.8
13.5
9.6
4.8
13.9
10.2
5.0
10.8
7.0
8.5
9.6
11.6
11.3
13.9
9.1
10.2
14.3
10.3
16.6
15.3
14.4
4.8
8.1
15.2
10.8
11.2
6.7
13.9
21.3
10.0
16.0
12.1
11.9
11.2
11.2
16.3
20.4
12.9
13.0
13.1
13.9
12.7

Fry
5
5
6
10
8
2
10
10
10
13
7
4
10
8
3
8
6
6
8
10
8
11
8
7
12
7
14
12
14
3
6
10
9
9
6
13
15
9
11
10
9
8
8
12
14
11
9
10
10
12

Gunning
5.08
4.85
6.41
5.67
6.72
6.81
6.41
6.07
8.48
6.07
6.15
4.09
6.57
6.40
3.05
7.03
4.11
4.26
5.94
5.26
5.82
7.81
5.19
7.09
9.33
8.19
9.05
9.34
8.23
3.25
5.47
9.61
7.16
5.80
5.81
7.35
13.48
7.20
11.00
8.83
8.24
7.83
6.95
11.01
9.78
8.13
8.99
8.95
9.72
6.06

Fog
6.0
5.9
8.4
8.1
10.0
9.3
10.0
10.0
13.0
10.0
8.7
4.8
10.0
9.4
3.2
10.0
5.2
5.5
8.1
8.1
9.4
13.0
7.2
9.7
15.0
12.0
17.0
16.0
14.0
3.5
7.5
16.0
11.0
8.5
7.5
13.0
25.0
11.0
17.0
14.0
13.0
11.0
12.0
21.0
19.0
13.0
16.0
14.0
15.0
9.7

GunningFog
22
24
28
32
36
31
38
43
42
42
32
22
40
35
20
35
27
27
30
37
40
43
31
31
47
37
54
49
49
21
30
46
37
34
27
46
64
38
46
40
40
36
44
59
58
42
50
40
43
40

Mean
12.82
10.32
9.82
9.27
9.24
8.38
8.10
8.07
7.11
7.06
6.78
6.72
6.54
6.15
5.74
5.41
5.34
5.10
4.96
4.88
4.81
4.80
4.78
4.63
4.53
4.39
4.35
4.08
4.01
3.94
3.87
3.77
3.74
3.69
3.62
3.58
3.57
3.48
3.39
3.30
3.19
3.13
2.98
2.82
2.55
2.42
2.30
2.13
2.09
1.60

12

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

13

The remaining columns in Table 3 show the readability estimates for each passage using the
Flesch, Flesch-Kincaid, Fry, Gunning, Fog, and Gunning-Fog indexes. All of these indexes,
except for the Gunning-Fog index, are meant to be on scales that represent grade levels in U.S.
schools. It is interesting that they are fairly comparable in some cases, thus indicating similar
relative difficulties for the passages. In other words, passages that appear to be relatively easy on
one index also tend to be relatively easy on the other ones as well, while passages that appear to
be relatively difficult on one index are also relatively difficult on the others.
Table 4
Correlation Coefficients L1 Readability Indexes and Russian Mean Difficulty 7

Flesch
Flesch-Kincaid
Fry
Gunning
Fog
Gunning-Fog
Russian Means

Flesch

FleschKincaid

Fry

Gunning

Fog

GunningFog

Means

1.00

.98

.92

.89

.93

.95

-.42

1.00

.90

.87

.92

.95

-.44

1.00

.70

.78

.88

-.41

1.00

.98

.87

-.45

1.00

.95

-.48

1.00

-.48
1.00

Table 4 shows the correlation coefficients for all possible pairs of the L1 readability
estimates in this study and the Russian means. The darker cells that spread across the table
diagonally indicate the correlations of each variable with itself, which in each case is of course
1.00. Note then that while a perfect correlation is 1.00 and total lack of correlation would equal
.00, all the values in the table are somewhere between the two extremes. Notice also that the
correlations in the last light gray column are negative values indicating that the means range
from easy to difficult in the opposite direction from the L1 readability indexes. In other words,
cloze passages with high means are relatively easy while those with low means are difficult. This
is the opposite (thus negative) from the readability indexes where high values indicate difficult
passages (suitable for higher grades) and low values indicate easy passages (suitable for lower
grades). It turns out that the L1 indexes are all moderately to highly correlated with each other
(between .70 and .98) depending on which one is examined. There also appears to be a relatively

Note that all of the correlation coefficients in Table 4 are statistically significant at p < .01 (one-tailed).

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

14

low degree of relationship between the Russian means and the various L1 indexes. Clearly, these
L1 indexes are more highly related to each other than they are to the Russian means.

DISCUSSION
The discussion will now return to the original research questions and will address each
question separately. The implications of these findings for Russian EFL readability estimations
will be explored in the Conclusions section.

1. Are Randomly Selected Cloze Tests Reliable and Valid Tools for Gathering Data on
Passage Difficulty?
In terms of reliability, Table 2 indicates that the cloze passages in this study are reasonably
reliable, though individual tests varied somewhat with Cronbach alpha reliability estimates
ranging from a moderate .646 to a relatively high .919. This means that these passages ranged
from being about two-thirds reliable (64.6%) to being more than nine-tenths reliable (91.9%). To
some degree, such variation in reliability can be related to the distributions of scores. The
magnitudes of the means (some of which were as low as 1.60) and standard deviations (many of
which were almost as large as their corresponding means) indicate that many of these
distributions were probably positively skeweda fact that would tend to depress the values of
Cronbach alpha. Nonetheless, these estimates represent the reliability of these cloze tests when
used under these conditions with these students.
In terms of validity, an argument can be built for the validity of the scores on these 50 cloze
tests as follows. Given that the cloze passages were constructed from randomly selected public
library books and that the items for the passages were semi-randomly selected (i.e., every 12th
word deletion), sampling theory would indicate that the passages form a representative sample of
the English language found in those library books. Therefore, it can be argued that the items
form a representative sample of the blanks that can be created from public library books. Given
that the validity of the scores from a set of items is defined as the degree to which they are
measuring what they purport to measure, the validity argument here is that these cloze items
have a high degree of content validity because they can be said to form representative samples of
the universe of all possible items (after Cronbach, 1970), if that universe is defined as blanks

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

15

created from the written language found in an American public library. [For much more on the
reliability and validity of these passages, see Brown, 1993; for more on test reliability and
validity issues, see Brown, 2005.]

2. To What Degree Are Traditional First Language Readability Indexes Related to the
Average Cloze Scores for the Same Passages (When Administered to Russian EFL Students)?
Tables 3 and 4 both show some degree of relationship between each of the L1 readability
indexes and the Russian means. Table 3 allows readers to actually inspect these relationships.
However, Table 4 shows the degree to which the L1 readability indexes are moderately to very
highly correlated with each other (ranging from .70 to .98). Thus, the L1 readability indexes
appear to be moderately to highly interrelated, which makes sense given that they are all based
on the same sorts of counts of syllables, words, sentences, etc. Table 4 also shows that L1
readability indexes are somewhat related to the performances of Russian EFL students as
indicated by the correlations of -.41 to -.48. Remember that these coefficients are on a scale from
no correlation (.00) to perfect correlation (1.00) and that they are negative values because the L1
readability and Russian mean scales indicate passage difficulty in opposite directions. Naturally,
all of this can be said to be true only for Russian university EFL students as sampled in this study
and for the cloze passages used here.

FUTURE DIRECTIONS

The results of this preliminary study indicate that there is certainly reason to move forward
with further analyses. It has shown that a variety of L1 readability indexes for this set of 50
passages are somewhat correlated with the average performances of Russian university students
on cloze versions of those same passages. This finding is important to think about. Why are the
L1 readability indexes only somewhat related to the Russian EFL means? This lack of
relationship could be due to any of the following: (a) that these L1 readability estimates are fine
indicators of passage readability for native speakers but not for Russian EFL learners; (b) that the
cloze passages are measuring something different from the simple readability measured by the
L1 indexes; (c) that the Russian EFL learners scores on the cloze passages are measuring

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

16

something much more complex than simple readabilitysomething like the students overall
proficiency levels rather than the reading difficulty of the passages.
We are hoping that the analyses that we are planning to perform on these data in the future
will help us to better understand these and other aspects of these cloze passages. In particular, the
directions we anticipate pursuing at the moment should lead us to answers to the following
research questions:
1. What other linguistic text variables (e.g., word length, word frequency for each blank, the
length of the sentence in which the blank is found, whether the word is of Germanic or
Latinate origin, etc.) should be included in such research?
2. How well do those linguistic text variables predict Russian EFL performances at the
passage level? At the item level? And in what combinations? [Using item-response
theory, factor analysis, multiple-regression analyses, and structural equation modeling
all of which will be explained in subsequent papers.]
3. What hierarchies of difficulty are found at the passage level for any of the linguistic
variables (separately or combined) that would have implications for second language
acquisition research? Similarly, what hierarchies of difficulty are found at the item level?
4. What differences and similarities would occur if the results of this study were compared
with the similar data gathered in Japan? With students from other language groups? With
students at other levels of study? Or at other ages?

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

17

REFERENCES

Alderson, J. C. (1978). A study of the cloze procedure with native and non-native speakers of
English. Unpublished doctoral dissertation, University of Edinburgh.
Alderson, J. C. (1979). Scoring procedures for use on cloze tests. In C. A. Yorio, K. Perkins, & J.
Schachter (Eds.) On TESOL 79 (pp. 193-205). Washington, DC: TESOL.
Bickley, A. C., Ellington, B. J., & Bickley, R. T. (1970). The cloze procedure: A conspectus.
Journal of Reading Behavior, 2, 232-234.
Bormuth, J. R. (1966). Readability: A new approach. Reading Research Quarterly, 1, 79-132.
Bormuth, J. R. (1967). Comparable cloze and multiple-choice comprehension tests scores.
Journal of Reading, 10, 291-299.
Brown, J. D. (1978). Correlational study of four methods for scoring cloze tests. Unpublished
MA thesis, University of California, Los Angeles.
Brown, J. D. (1980). Relative merits of four methods for scoring cloze tests. Modern Language
Journal, 64, 311-317.
Brown, J. D. (1988). Tailored cloze: Improved with classical item analysis techniques. Language
Testing, 5, 19-31.
Brown, J. D. (1989). Cloze item difficulty. JALT Journal, 11, 46-67.
Brown, J. D. (1998). An EFL readability index. JALT Journal, 20(2), 7-36.
Brown, J. D., Chen, Y., & Wang, Y. (1984). An evaluation of native-speaker self-access reading
materials in an EFL setting. RELC Journal, 15, 75-84.
Carrell, P. L. (1987). Readability in ESL. Reading in a Foreign Language, 4, 21-40.
Chall, J. S. (1958). Readability: An appraisal of research and application. Columbus, Ohio:
Ohio State University.
Cronbach, L. J. (1970). Essentials of psychological testing (pp. 145-146). New York: Harper &
Row.
Flesch, R. (1948). A new readability yardstick. Journal of Applied Psychology, 32, 221-233.
Fry, E. B. (1977). Frys readability graph: Clarifications, validity, and extension to level 17.
Journal of Reading, 21, 242-252.
Fry, E. B. (1985). The NEW reading teachers book of lists. Englewood Cliffs, NJ: Prentice-Hall.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

18

Gilliam, B., Pea, S. C., & Mountain, L. (1980). The Fry graph applied to Spanish readability.
The Reading Teacher, 33, 426-430.
Greenfield, G. (1999). Classic readability formulas in an EFL context: Are they valid for
Japanese speakers? Ed.D. dissertation, Temple University.
Greenfield, J. (2004). Readability formulas for EFL. JALT Journal, 26(1), 5-24.
Hamsik, M. J. (1984). Reading, readability, and the ESL reader. Unpublished doctoral
dissertation, Florida State University, Tallahassee, FL.
Haskell, J. (1973). Refining the cloze testing and scoring procedures for use with ESL students.
Unpublished doctoral dissertation, Columbia University, New York.
Klare, G. R. (1963). The measurement of readability. Ames, IA: Iowa State University.
Klare, G. R. (1984). Readability. In R. Barr, M. L. Kamil, & P. Mosenthal (Eds.), Handbook of
reading research (pp. 681-744). New York: Longman.
Larson, R. (1987). How Readability was created. In Scandinavian PC Systems. Readability
program for the IBM PC, XT and AT (pp. 8-1 to 8-20). Rockville, MD: Scandinavian PC
Systems.
Miller, G. R., & Coleman, E. B. (1967). A set of thirty-six prose passages calibrated for
complexity. Journal of Verbal Learning and Verbal Behavior, 6, 851-854.
Moyle, D. (1970). Readabilitythe use of cloze procedure. In J. Merrit (Ed.), Reading and the
curriculum (pp. 159-168). London: Ward Lock Educational.
Nguyen, L. T., & Henkin, A. B. (1982). A readability formula for Vietnamese. Journal of
Reading, 26, 243-251.
Ransom, P. E. (1968). Determining reading levels of elementary school children by cloze testing.
In J. A. Figurel (Ed.), Forging ahead in reading (pp. 477-482). Newark, DE: International
Reading Association.
Taylor, W. L. (1953). Cloze procedure: A new tool for measuring readability. Journalism
Quarterly, 30, 414-438.
Taylor, W. L. (1957). Cloze readability scores as indices of individual differences in
comprehension and aptitude. Journal of Applied Psychology, 41, 19-26.
Zakaluk, B. L., & Samuels, S. J. (1988). Readability: Its past, present, and future. Newark, DE:
International Reading Association.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

19

Zamanian, M., & Heydari, P. (2012). Readability of texts: State of the art. Theory and Practice
in Language Studies, 2(1), 43-53.

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

20

APPENDIX A: PARTICIPATING UNIVERSITIES


We owe a great debt of gratitude to our colleagues at the following 38 institutions who helped us
gather the data upon which this study is based. Without their cooperation and help, and the
efforts of their students, this research project would simply not exist.

1. Chelyabinsk Law Institute


2. Chelyabinsk State University
3. International Market Institute (Samara)
4. Kazan branch of the Russian International Academy for Tourism
5. Kazan Military Institute
6. Kazan State Technical University
7. Kolomna State Pedagogical University
8. Krasnodar State University
9. Mordovian State University
10. Moscow State University
11. Novocherkassk Polytechnic Institute
12. Novosibirsk State University
13. Orenburg State University
14. Presidential Cadet College (Orenburg)
15. Rostov/Don Institute of Management, Business and Law
16. Rostov/Don State University
17. Ryazan State University
18. Samara Aerospace university
19. Samara State Academy of Social Sciences and Humanities
20. Samara State University
21. Samara State University of Architecture and Civil Engineering
22. Saratov State Pedagogical University
23. Saratov State University
24. Smolensk University for the Humanities
25. Solykamsk State Pedagogical University
26. South-Ural State University

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

27. St. Petersburg State University


28. Surgut State University
29. Syktyvkar State University
30. Syzran branch of Samara State Technical University
31. Taganrog Institute of Management and Economics
32. Taganrog State Pedagogical University
33. Togliatti Academy of Management
34. Tomsk Polytechnic University
35. Ulyanovsk State University Institute for International Relations
36. Volga State University of Technology (former Mari State Technical University)
37. Voronezh State University
38. Voronezh State University of Architecture and Civil Engineering

21

BROWN ET AL. - CLOZE PROCEDURE AS A TOOL FOR ESTIMATING READABILITY

22

APPENDIX B: EXAMPLE CLOZE PASSAGE


(ADAPTED FROM BROWN, 1989)
Name________________________________ Native Language____________________
(Last)
(First)
Sex__________________ Age___________ Country of Passport________________
DIRECTIONS:
1. Read the passage quickly to get the general meaning.
2. Write only one word in each blank. Contractions (example: dont) and possessives (Johns
bicycle) are one word.
3. Check your answers.
NOTE: Spelling will not count against you as long as the scorer can read the word.

EXAMPLE: The boy walked up the street. He stepped on a piece of ice. He fell
(1)____________, but he didnt hurt himself.

A FATHER AND SON


Michael Beal was just out of the service. His father had helped him get his job at Western.
The (1)____________ few weeks Mike and his father had lunch together almost every
(2)____________ . Mike talked a lot about his father. He was worried about (3)____________
hard he was working, holding down two jobs.
You know, Mike (4)____________ , before I went in the service my father could do just
(5)____________ anything. But hes really kind of tired these days. Working two
(6)____________ takes a lot out of him. He doesnt have as much (7)____________ . I tell him
that he should stop the second job, but (8)____________ wont listen.
During a smoking break, Mike introduced me to his (9)____________ . Bill mentioned that
he had four children. I casually remarked that (10)____________ hoped the others were better
than Mike. He took my joking (11)____________ and, putting his arm on Mikes shoulder, he
said, Ill be (12)____________ if they turn out as well as Mike.
(continues ...)

You might also like