**2.4 Analysis**

As the number of items in the achievement tests was not equal among different graders, the correct response rate for each grade was calculated as a proportion and hence the arc sine root transformation was applied to the correct response rates (Sheskin 2007). Since the number of children with developmental disorders was limited at each grade, we abandoned the analysis of variance using two levels of each independent variable. Instead, we created new variables using the four categories of items mentioned above.

Specifically, in order to examine the effect of orthographical demand (i.e. visual complexity), we averaged the correct response rate for (i) characters that consist of a small number of strokes and have only one pronunciation and (iii) characters that consist of a small number of strokes and have more than one pronunciation on one hand to create a variable representing the participants' performance for visually less complex characters, and averaged the correct response rate for (ii) characters that consist of many strokes and have only one pronunciation and (iv) characters that consist of many strokes and have more than one pronunciation on the other hand to create a variable representing the participants' performance for visually more complex characters. Two more variables representing their performance for phonologically less complex characters and their performance for phonologically more complex characters were created in a similar manner.

The mean correct response rate for typically developing children at each grade was further analyzed using the statistical tests according to its distribution after the test for the homogeneity of variance and the test for the normality of distribution. Since the mean correct response rates were not significantly different between the gender groups of typically developing children at each grade according to the Mann Whitney U test, we calculated the mean and the standard deviation of the entire group, containing both girls and boys. Using this mean and the standard deviation as the basis, we then calculated the Zscores for children with developmental disorders at each grade, although the proportion of males and females was different between typically developing children and children with developmental disorders.

#### **3. Results**

The result of the experiment seems to support the following three statements:

Assessing Orthographical and Phonological Impairments 75

Fig. 2. Overall correct response rate of writing by typically developing children.

Fig. 3 represents the Z scores for the correct response rate of both dyslexics and ADHD patients for the reading and writing achievement tests. The left side column shows the Z scores of the reading tests and the right side column shows those of the writing tests. Each row shows the Z scores of each grade. Filled data points in black represent the cases with

From those graphs we see that Z scores for writing tend to be low with both developmental disorders, that dyslexics' reading scores likewise tend to be low, but that ADHD patients' Z scores for reading tend to be close to normal (around -1.0 SD or above), suggesting relatively

The pattern of overall performance alone does not necessarily allow us to distinguish ADHD patients and dyslexics. In other words, some ADHD patients and some dyslexics show an indistinguishable pattern of overall performance, as we see with Case 4 (dyslexic) and Case 13 (ADHD) in Grade 3 and Case 12 (dyslexic), Case 16 (ADHD) and Case 17 (ADHD) in Grade 5. Thus, in order to distinguish the two types of disorders on the basis of reading and writing achievement tests, it is necessary to examine not only the overall performance of the children,

Fig. 4 reports the typically developing children at each grade's mean correct response rate for reading achievement tests as a function of orthographical complexity. Error bars indicate standard deviations. There was significant difference between the mean correct response rate for orthographically complex characters and that for orthographically simple characters at Grade 2. The correct response rate was higher when the character was orthographically complex than when the character was orthographically less complex (Z=2.51, *p*<.01, *r*=.27). The sampling distribution was similar between the levels of orthographical complexity in

**3.1.2 Overall performance by children with developmental disorders** 

dyslexia whereas unfilled data points represent the cases with ADHD.

but also their performance for each of the four (two by two) question types.

**3.2.1 Orthographical performance by typically developing children** 

**3.2 Performance as a function of orthographical complexity** 

other grades, according to Wilcoxon's signed rank test.

minor impairment of reading abilities.


#### **3.1 Overall performance**

#### **3.1.1 Overall performance by typically developing children**

Fig. 1 reports the mean correct response rate of typically developing children at each grade for the reading achievement tests. Error bars indicate one standard deviation of uncertainty. The sampling distribution was similar among different grades, according to the Kruskal-Wallis test (*p* = .1).

Fig. 1. Overall correct response rate of reading by typically developing children.

Fig. 2 represents the means of the correct response rate of typically developing children at each grade for the entire writing achievement tests. Error bars indicate one standard deviation over sampling distribution. The sampling distribution was significantly different among different grades, according to the Kruskal-Wallis test (p < .0001). The post hoc tests revealed that the mean correct response rate of Grade 4 and 5 was significantly lower than that of Grade 3 (U=3830.5, Z=-3.22, *r*=-.23; U=3652.0, Z=-4.07, *r*=-.28, respectively) after Bonferroni corrections.

First, while children with dyslexia had trouble both with writing and with reading, children with ADHD had trouble mainly with writing and not necessarily with

Second, compared to children with dyslexia, children with ADHD had more trouble writing Chinese characters which consist of many strokes and thus are visually more complex, although both children with ADHD and dyslexia seemed to have less trouble

Third, compared to children with ADHD, dyslexic children had more trouble reading (if not writing) Chinese characters which have more than one possible pronunciation

Fig. 1 reports the mean correct response rate of typically developing children at each grade for the reading achievement tests. Error bars indicate one standard deviation of uncertainty. The sampling distribution was similar among different grades, according to the Kruskal-

Fig. 1. Overall correct response rate of reading by typically developing children.

Fig. 2 represents the means of the correct response rate of typically developing children at each grade for the entire writing achievement tests. Error bars indicate one standard deviation over sampling distribution. The sampling distribution was significantly different among different grades, according to the Kruskal-Wallis test (p < .0001). The post hoc tests revealed that the mean correct response rate of Grade 4 and 5 was significantly lower than that of Grade 3 (U=3830.5, Z=-3.22, *r*=-.23; U=3652.0, Z=-4.07, *r*=-.28, respectively) after

a. Overall performance

b. Orthographic performance

c. Phonological performance

**3.1 Overall performance** 

Wallis test (*p* = .1).

Bonferroni corrections.

reading visually more complex characters.

and are thus arguably phonologically more complex.

**3.1.1 Overall performance by typically developing children** 

reading.

Fig. 2. Overall correct response rate of writing by typically developing children.

## **3.1.2 Overall performance by children with developmental disorders**

Fig. 3 represents the Z scores for the correct response rate of both dyslexics and ADHD patients for the reading and writing achievement tests. The left side column shows the Z scores of the reading tests and the right side column shows those of the writing tests. Each row shows the Z scores of each grade. Filled data points in black represent the cases with dyslexia whereas unfilled data points represent the cases with ADHD.

From those graphs we see that Z scores for writing tend to be low with both developmental disorders, that dyslexics' reading scores likewise tend to be low, but that ADHD patients' Z scores for reading tend to be close to normal (around -1.0 SD or above), suggesting relatively minor impairment of reading abilities.

The pattern of overall performance alone does not necessarily allow us to distinguish ADHD patients and dyslexics. In other words, some ADHD patients and some dyslexics show an indistinguishable pattern of overall performance, as we see with Case 4 (dyslexic) and Case 13 (ADHD) in Grade 3 and Case 12 (dyslexic), Case 16 (ADHD) and Case 17 (ADHD) in Grade 5. Thus, in order to distinguish the two types of disorders on the basis of reading and writing achievement tests, it is necessary to examine not only the overall performance of the children, but also their performance for each of the four (two by two) question types.
