主题：【原创】浅谈汉字与拼音文字的比较 （上） -- 人间树

RESEARCH DESIGN

In order to make sure that the reading speed and reading comprehension rate are the function of orthography, extraneous variables should be strictly controlled for during the test. To do this, an experimental design was called for. A survey design using large random samples was incorporated with the experimental tests to make the findings more reliable and valid.

SUBJECTS

College freshman students were randomly selected from Capital Normal University in China and Brigham Young University in Utah to represent respectively native Chinese readers and native English readers. The sample size for each language group was about 140, with approximately equal numbers of males and females from various areas of study. College freshmen were chosen as research subjects out of the following three reasons:

1. with the same educational level, they should be comparable in reading ability,

2. college students are generally more careful and conscientious readers than non-student populations, and

3 in the higher education environment, the procedure of random sampling is easier than in others.

Capital Normal University and Brigham Young University were selected as the sampling frames from the two populations because they are believed to be comparative in their freshmen students' reading abilities.

MATERIALS

Eight English passages were selected from the Test of English as a Foreign Language (TOEFL), and eight from the Chinese Standard Test (HSK: Hanyu Shuiping Kaoshi). Each of the sixteen passages is followed by two or more multiple-choice questions measuring reading comprehension. The English passages and questions were translated into Chinese, and the Chinese ones into English. In order to keep translation errors to a minimum, back translation techniques were used with the help of several bilingual scholars. Thus, each of the two groups was tested on sixteen passages in their specific form. The total number of questions for the sixteen passages is 60. The contents of all the passages are either of popular science in nature or basic social issues. As a principle of TOEFL and HSK, all the passages are neutral in terms of points of view and unbiased against any group of people.

In order to reduce the influence of extraneous variables during the test, each of the 16 passages is printed on a separate page, with its corresponding questions on the back of the same page. Respondents were instructed not to re-read the passage once they had finished reading it.

In the test for both languages, the first passage is originally from the Test of English as a Foreign Language (TOEFL), and the second is from the Chinese Standard Test (HSK: Hanyu Shuiping Kaoshi). All eight English and Chinese passages are alternatively ordered so that the results will not be biased to readers of any language, even if a subject turns in an incomplete test.

The English version of the test was printed in WordPerfect default font size, and the Chinese version was printed in Nan Ji Star default size. Both sizes are comparable and regular and are not supposed to create difficulties or advantages to a reader of any version. The English and Chinese versions of the test are available for inspection upon request from interested readers.

MEASUREMENTS

The major independent variable in the study is orthography. This is a dummy variable varying from English to Chinese.

Reading speed and comprehension rate are two major dependent variables. Reading speed is measured by the number of minutes in which a subject finishes reading all 16 passages and completing all 60 multiple choice questions. Comprehension rate is measured by the number of questions answered correctly by a subject.

A third dependent variable measuring reading efficiency is actually an index of reading speed and comprehension rate. Efficiency is a ratio of comprehension (number of questions answered correctly) to the speed (number of minutes used to finish the test):

Efficiency = Correct Items / Minutes Used

A high efficiency rate indicates either more "correct items" or less "minutes used" or both. For our test material, the value of "correct items" should vary from 0 to 60, and the value of "minutes used" should theoretically vary from 1 to infinity. However, the test was designed for a normal native reader of either English or Chinese to finish with adequate comprehension within 30 minutes. Pilot studies conducted in both English and Chinese languages indicated that the majority of readers could finish the whole thing within about 40 minutes. Subjects of the study were asked to read the passages in one of the two language forms (English and Chinese) and then answered multiple-choice questions measuring their reading comprehension. All subjects were required to treat the reading as a timed test and encouraged to finish it as quickly and correctly as possible. The test administrators supervised the test and recorded the time (minutes) used by each subject to finish the test. The comprehension scores (number of questions answered correctly) would be recorded after the tests were collected.

ANALYSES AND FINDINGS

Subjects for the research project are from Capital Normal University in Beijing, China and Brigham Young University in Provo, Utah, USA. They are representing two different languages under study: Chinese and English. The two institutions were selected both for comparability and for convenience. They are of similar level in terms of quality of students in each country.

For each sample, the test was administered to about 130 students. Since we prefer to study only freshmen students, some non-freshmen cases were deleted from the samples. After removing the cases that were far from complete, we have 126 subjects for the Chinese sample, and 111 for the English sample. The demographic characteristics of the two samples are shown in Table 1.

For the current data, reliability tests were performed on mean scores of all the 16 passages. Three scales were tested for the instrument's reliability: (1) for all the 16 passages, (2) for the 8 odd-numbered passages, and (3) for 5 randomly selected passages. Table 2 presents the alpha value and other information for each scale tested. As we can see from the table, the more items involved in the reliability test, the higher the alpha tends to be. Generally, the instrument is adequately reliable as a test for reading comprehension.

Table 2 also presents the results of reliability tests for each of the two samples. Comparing the alpha values between the English and Chinese versions of the instrument, we find that the English version is more reliable than the Chinese version for the scales with 8 items and 5 items, but not for the scale with all 16 items. Generally, the versions are not different in their reliability levels.

As standard foreign language tests, both TOEFL and HSK have endured long-time examinations and refinements for their reliability and validity. For a validity test, we are looking at the degree to which the tests are measuring what they are supposed to measure. Both tests are aimed at a valid test of an examinee's proficiency of a foreign language. In the reading comprehension sections from which passages were selected for the current project's use, both TOEFL and HSK were designed to test an examinee's reading speed and comprehension. Reading speed and comprehension are what we are testing our subjects on. The only difference between the original purpose of the TOEFL/HSK and the purpose in the current project lies in the subjects: their examinees are foreign language readers, but ours are native readers of the language. For native readers, the passages should be much easier, and they can finish reading the whole thing in a shorter time than non-native readers. Therefore, 16 passages were put together to form an adequately long test so that variance in reading speed and comprehension can be detected for native readers.

The hypotheses specified in the research have been tested one after another. As a result, the data supported all the five hypotheses.

Hypothesis #1: Native Chinese reading is generally faster than native English reading. To compare the difference in reading speed between the English native readers and Chinese native readers, a one-way Analysis of Variance (ANOVA) was performed with the number of minutes used to finish the whole reading as the dependent variable and the group (with two values: English and Chinese) as the dependent variable. The mean number of minutes used by Chinese readers to finish the test was 24.7 (S.D. = 4.7), while that by English readers was 26.6 (S.D. = 5.1), with df=1, F=8.6, and significant level=.004. Chinese read about 2 minutes faster and with a smaller standard deviation than the English readers do on the test.

Hypothesis #2: On the reading test, native Chinese readers score higher than native English readers in comprehension rates for the same reading contents. When reading comprehension is measured by the number of questions answered by subjects correctly, we can see which group understands the same material with more comprehension of the texts. The mean score for the Chinese group is 50.1 (S.D.=6.0), and the mean score for the English group is 43.9 (S.D. = 6.8), with df=1, F=54.7, and significant level=.000. On average, the Chinese readers can answer 6 more questions correctly than the English readers on a 60-question test.

Hypothesis #3: When reading efficiency is calculated by comprehension score (number of questions answered correctly) divided by speed (number of minutes used to finish the reading), native Chinese reading is more efficient than native English reading. A fast reader may not have a good comprehension score, and inversely, a reader who has scored high on comprehension may have done so at the expense of time. The new variable of efficiency was created to balance out the discrepancies. As we already know, the efficiency score is obtained by the number of correct answers divided by the number of minutes used to finish the test. The mean efficiency score for the Chinese readers is 2.10 (S.D.=.45) and for the English readers, 1.70 (S.D.=.39), with df=1, F=50.6, and significant level=.000.

Hypothesis #4: With comprehension rates being controlled for, native Chinese reading is generally faster than native English reading. As discussed earlier, a fast reader may not score high on the comprehension test, and a reader who has a high comprehension score may do so at expense of the time. A correlation test for the two variables was performed to address the concern. The result is that the correlation coefficient is .002 and the probability level is .97. There is almost no relationship between the number of minutes used and the number of questions answered correctly; a person who has spent more time on the reading does not necessarily score high on the comprehension test.

To further address the concern, a multiple regression model is designed:

Speed = a + b[sub1](Orthography) + b[sub2](Comprehension)

where the "speed" (number of minutes used to finish the test) is determined by "a" (an intercept or constant value for the model) and "b[sub1]" (the slope provided by the model for the variable of orthography) times the value of orthography, and additionally "b[sub2]" times the comprehension score. In this model, the speed is predicted with orthography (English or Chinese) or comprehension held constant. In other words, when we determine a subject's reading speed with a known orthography value, the comprehension is held constant, and when we determine his reading speed with a known comprehension score, the orthography is held constant.

As can be seen from Table 3, the Y intercept (constant) is 23.39. This is about the average speed for everyone in the samples to finish the test. However, the actual speed of a certain reader is predicted by his/her orthography (English or Chinese) and comprehension score as follows:

Speed = 23.39 + (-2.32)(Orthography) + (.07)(Comprehension).

Since the variable "orthography" (or sample) is recoded as 0 = English and 1 = Chinese, a negative slope value for orthography for the model indicates that the time value is always lower for the Chinese reader than for an English reader. In this case, the speed of reading for the Chinese is always faster than that for the English readers by 2.32 minutes. The comprehension slope value in the table tells us that with every point made in comprehension (every question answered right), the time is increased by .07 minutes. However, the relation is too weak to be significant at the .05 probability level. With the strong predicting power from the orthography, the multiple regression model has an F value of 5.36 and is significant at .005 level.

Hypothesis #5: With speed being controlled for, the comprehension rate in native Chinese reading is higher than that in native English reading. Table 3 also has the information of the multiple regression model that predicts comprehension with a known value of orthography and speed. The second model has comprehension as its dependent variable:

Comprehension = a + b[sub1](Orthography) + b[sub2](Speed)

Again, in this model the comprehension value is predicted with orthography (English or Chinese) or speed held constant. In other words, when we determine a subject's comprehension score with a known orthography value, the speed is held constant, and when we determine his comprehension with a known speed, the orthography is held constant.

For the model, the Y intercept of 40.68 as shown in Table 3 indicates that the minimum comprehension mean score is about 40.68. The actual score of comprehension can be predicted by orthography and speed as follows:

Comprehension = 40.68 + (6.39)(Orthography) + (.12)(Speed).

Since English has been coded as 0 and Chinese as 1, this multiple regression model tells us that a Chinese reader always scores at least 6.39 more than a English reader the comprehension. With orthography being held constant, every one minute used in the test will help an average subject increase his/her comprehension score by .12, although the relationship is not significant at .05 level. Again, with the strong predicting power from the orthography, the multiple regression model has an F value of 28.55 and a probability level lower than .0001.

Both models with either speed or comprehension as dependent variable have supported the general hypothesis of the study that the orthography (English or Chinese) makes difference in reading speed and comprehension. Chinese readers read faster than the English readers, and Chinese readers read the passages with less comprehension errors than their English counterparts