**3. Results**

184 Learning Disabilities

The study presented here, which is a descriptive study, consisted of two comparative and related samples, chosen simultaneously from 30 schools. The samples were selected to complement each other. In fact, the research team was carrying out a series of studies in the province of León which required samples of students from three different categories which are matched for general characteristics: a group of students with LD, another group without

The sample of the teachers was selected on the basis of the criterion explained above. They were teachers who were responsible for students with and without learning difficulties and/or low achievement in infant and early primary education. The next step was the fieldwork itself. It consisted, firstly, of establishing telephone contact with the school principals to obtain permission to visit and carry out the protocol with the language teachers of the selected school years in each of these schools. Two researchers then visited the schools on the agreed dates and times and asked teachers to complete the protocols. The sample selection was performed directly by the two researchers and not by other means, to ensure the reliability and validity of the instruments in the collection of information. Data collection

The procedure followed for the data collection from students can be summarized in the following steps. Firstly, general measures of sample selection were applied. Different members of the research team, with an approximate duration of 10 to 20 minutes, carried out an IQ measurement using the Cattell assessment in small groups. After the Cattell assessment, students were asked to carry out different written composition tasks. Small groups of students carried out individual free text exercises (García, Marbán, and de Caso, 2001), to determine their level of achievement in writing. Several comparison-contrast texts were written to obtain the indicators and the measures of the product of the writing (text and reader-based measures). These measures were used to relate the students' achievement in writing to the role of the teachers' practice, which constitutes the object of this study. Our basis for this sample came from an initial sample of more than 350 students from previous studies carried out by the research team, of which only 111 students were selected for this study. Simultaneously, the teachers were asked to complete evaluation scales regarding the each student's general achievement . This task took some time and once finished the forms

Along with the application of these instruments and tasks, a further set of tests and questionnaires were applied. These were related to attention, working memory, and the study of the online processes used in written composition. They were measured by means of a writing log, but they are not included in this study, as they are part of an overarching project including different studies with broader goals than those presented in this article. Once all the assessments were carried out in the 30 schools, the members of the research team analyzed all the texts. Members of the team, who had received specific training over several meetings, including the study of the correction dossier and the systematic supervision of the written texts, carried out their meticulous correction. The texts were corrected twice and anonymously with the purpose of reaching an agreement between coders. The approximate time dedicated to the correction of texts for each student was of two hours, including the codification and computerization of the data. The correction was

**Procedure** 

were collected.

LD and a group of students with ADHD.

was conducted over a period of five months.

After the creation of a single matrix on SPSS (v. 13.0) with the variables generated for the data regarding teachers and students, the following types of analysis were carried out: multivariate analyses of variance (MANOVAS) using the SPSS Model Lineal General module (MLG) and multiple linear regression analyses (ARL).

#### **General Linear model**

The multivariate contrasts in the student based measures based on their typology (ADHD, LD versus without LD), indicate highly significant statistically results with a large size effect [F (84, 28) = 3.52; p = .001; ŋ ² = .914].

For the execution of these multivariate analyses of the three student typologies (ADHD, with and without LD) and, as dependent variables, the different measures obtained for the students and the teachers were taken as intersubject factors. The purpose of this was to highlight the differences between the teachers as regards the role of practice in the teaching of writing. We also sought to show the differences according to student typology and to try to extract some pattern to help understand the relationship between the teachers' writing teaching practice and the students' achievement.

The tests concerning the intersubject/intergroup effects also indicated statistically significant results in most of the dependent variables related to students and to teachers. The size effects, in general, were large. When the grouping variable (intersubject factor or fixed factor) belonging to a specific group was taken (with ADHD, with and without LD), statistically significant differences were observed, both in the writing tasks and in comparison–contrast texts. In the comparison-contrast texts significant differences were observed for the variables of information generation, organization of ideas, structure, reader-based quality measures, etc. For example, for organization of ideas in relational coherence (p = .001; ŋ ² =.217); or for reader-based measurement as concerns structure (p= .001; ŋ ² =.213); and for reader-based evaluation (order-quality) (p = .001; ŋ ² =.284). Also, in the writing task, statistically significant differences were observed for idea generation and total productivity (p = .001; ŋ ² =.268); in the reader based evaluation (order and structure) (p = .001; ŋ ² =.486); and in quality (p = .001; ŋ ² =.383).

As concerns what teachers actually did (see Table 5), it was interesting to observe that it differed according to the students taught, in motivation (close to statistical significance), in natural teaching approach, and in self-efficacy (very close to statistical significance). The tests of the intersubject effects indicated statistical significance for the opinion variables,

The Quality of Teaching Determines Students' Achievement in Writing 187

Variable LD vs. WLD LD vs. ADHD WLD vs. ADHD

Total Productivity .001 n. s. .047 Organization of ideas: consistency referential .051 n. s. n. s. Organization of ideas: total coherence .049 n. s. n. s. Other aspects: overall coherence .041 n. s. n. s. Reader based evaluation: amount, structure .001 .056 .001 Reader based evaluation: summary, consistent .001 n. s. .001 Reader based evaluation: amount, quality .001 n. s. .002 Reader based evaluation: order, structure .001 n. s. .001 Reader based evaluation: order, coherence .001 n. s. .001 Reader based evaluation: order, quality .001 n. s. .001

 *Comparison-contrast (p)* 

Productivity total n. s. n. s. .075 Relational Consistency .029 n. s. .002 Total Consistency n. s. n. s. .052 Other aspects: overall coherence n. s. .039 .001 Reader based evaluation: amount, structure .005 n. s. .055 Reader based evaluation, summary: consistent .005 n. s. .001 Reader based evaluation, amount: quality .003 n. s. .002 Reader based evaluation, order: structure .008 n. s. .005 Reader based evaluation, order: coherence n. s. n. s. .027 Reader based evaluation, amount: quality .001 n. s. .001

Table 6. Post-hoc significant contrasts in the multivariate analyses, both for the students'

Fig. 1. Statistically significant differences in the teachers' variables, depending on student

**Natural Theor Approach**

**Classroom: Skills**

**General Self-efficacy**

**Without LD LD AHDD**

typology.

**O. Motivation**

**O. Planning/Review**

**O. Practice**

**O. Family**

**Formal Theor Approach**

writing tasks (by type) and for the teachers' actions (by type of students)

*Essay (p)* 


motivation subcomponent (p = .073; ŋ ² =, 091); approach, natural learning subcomponent (p = .002; ŋ ² =, 205); and, close to statistical significance for general self-efficacy (p = .058; ŋ ² =, 098).

Table 5. Statistically significant results: multivariate analysis of variance in the teachers' measures (PRAES) between groups of students with ADHD, DA and without DA

When the contrasts between the significant variables obtained in the tests of the intersubject effects (between the groups) were collated post hoc, we found statistically significant differences in many of them (see Table 6). For example, there were significant differences among the post contrasts and between the group with LD and the group without LD. The same was noted for the ADHD group, that showed significant differences compared to the group without LD. However, no significant differences were noted between the groups of students with LD and those with ADHD. This pattern was observed in the variables that refer to the writing task in total productivity (LD as opposed to without LD, p = .001; without LD as opposed to ADHD, p =, 047) and for organization of ideas, referential coherence (with as opposed to without LD, p =,051). Other patterns noted are also of interest, for example, the significant differences between the three groups, as is the case with the reader-based structure measures (with LD as opposed to without LD, p = .001; LD compared with ADHD, p = .056; and without LD compared with ADHD, p =,001).

Also, we would also like to highlight the results concerning the teachers – the statistical significance related to the motivation variable (p = .073), natural teaching approach (p =.002) and general self-efficacy (p = .058).

The post hoc contrasts indicated different patterns between the students with and without LD and ADHD with regard to the teacher related variables. There were likewise differences between the group with LD and those without LD. The ADHD group is significantly different from the group without LD, but this is not so between the LD and ADHD groups. This pattern is observed in the natural teaching approach variable (LD as opposed to ADHD, p = .002; and without LD as opposed to ADHD, p =, 072). The non-significant variables related to the PRAES include those which correspond to: Opinion and the subcomponents of practice (p = .109), family (p = .170) and teacher training (p = .650); the formal approach (p = .315); the actual classroom behaviors and the subcomponents of abilities (p = .410), materials (p = .874), procedures (p = .271), texts(p = .278) activities (p = . 454); and personal self-efficacy (p = .913). These variables, to some extent, predict teachers' behaviors in the teaching of written composition (see Figure 1).

motivation subcomponent (p = .073; ŋ ² =, 091); approach, natural learning subcomponent (p = .002; ŋ ² =, 205); and, close to statistical significance for general self-efficacy (p = .058; ŋ ² =, 098).

Variables M α M Α M α F p ŋ² Opinion: Reasons 29.59 3.84 28.18 4.96 31.32 3.72 2.74 .073 .091 Opinion: Planning and revision 24.12 2.08 24.91 2.22 26.16 1.97 1.18 .314 .041 Opinion: Practice 30.59 4.34 31.64 3.23 33.00 2.47 2.30 .109 .077 Opinion: Family 15.41 2.47 14.09 2.91 15.26 1.55 1.83 .170 .062 Formal Education Approach 39.88 8.08 37.64 7.65 41.21 6.90 1.18 .315 .041 Natural Learning Approach 34.59 7.06 27.91 4.63 30.26 4.86 7.08 .002 .205 Classroom: Skills 24.29 3.72 23.05 3.33 22.95 2.97 .907 .410 .032 General Self-efficacy 33.05 10.06 38.31 4.79 37.00 4.92 3.00 .058 .098

Table 5. Statistically significant results: multivariate analysis of variance in the teachers' measures (PRAES) between groups of students with ADHD, DA and without DA

compared with ADHD, p = .056; and without LD compared with ADHD, p =,001).

behaviors in the teaching of written composition (see Figure 1).

and general self-efficacy (p = .058).

Also, we would also like to highlight the results concerning the teachers – the statistical significance related to the motivation variable (p = .073), natural teaching approach (p =.002)

The post hoc contrasts indicated different patterns between the students with and without LD and ADHD with regard to the teacher related variables. There were likewise differences between the group with LD and those without LD. The ADHD group is significantly different from the group without LD, but this is not so between the LD and ADHD groups. This pattern is observed in the natural teaching approach variable (LD as opposed to ADHD, p = .002; and without LD as opposed to ADHD, p =, 072). The non-significant variables related to the PRAES include those which correspond to: Opinion and the subcomponents of practice (p = .109), family (p = .170) and teacher training (p = .650); the formal approach (p = .315); the actual classroom behaviors and the subcomponents of abilities (p = .410), materials (p = .874), procedures (p = .271), texts(p = .278) activities (p = . 454); and personal self-efficacy (p = .913). These variables, to some extent, predict teachers'

When the contrasts between the significant variables obtained in the tests of the intersubject effects (between the groups) were collated post hoc, we found statistically significant differences in many of them (see Table 6). For example, there were significant differences among the post contrasts and between the group with LD and the group without LD. The same was noted for the ADHD group, that showed significant differences compared to the group without LD. However, no significant differences were noted between the groups of students with LD and those with ADHD. This pattern was observed in the variables that refer to the writing task in total productivity (LD as opposed to without LD, p = .001; without LD as opposed to ADHD, p =, 047) and for organization of ideas, referential coherence (with as opposed to without LD, p =,051). Other patterns noted are also of interest, for example, the significant differences between the three groups, as is the case with the reader-based structure measures (with LD as opposed to without LD, p = .001; LD

Measure/group ADHD (N = 40) DA (N = 35) SDA (N = 36)


Table 6. Post-hoc significant contrasts in the multivariate analyses, both for the students' writing tasks (by type) and for the teachers' actions (by type of students)

Fig. 1. Statistically significant differences in the teachers' variables, depending on student typology.

The Quality of Teaching Determines Students' Achievement in Writing 189

carried out and a corrected R² of .067 was obtained. This is significant, although it only generates a regression model or equation with the predicting variables of PRAES for opinion with the family subcomponent [β = .276; t = 2.817; p = .006] and for approach, with the subcomponent of formal approach [β = .217; t = 2.213; p = .029]. It does not generate any

In summary, regression analysis provides statistically significant data for predicting all the writing variables within PRAES, which presents interesting data regarding predictive validity. It also indicates the great predictive potential of the instruments applied, as they

The objective of the present study was to consider the teachers' self-regulation in the teaching of written composition in relation to the achievement of students with and without LD and/or under achievement. It was expected that the results of this study would show that the teachers' beliefs both affect and have a strong influence in their classroom practices concerning written composition and also that this predicts students' success. According to the results obtained it is possible to affirm that, broadly speaking, the objective was achieved. As for the hypotheses, they were only proven for some measures but not for others. For example, data was obtained which supports the differences between the students based on the PRAES assessment. The potential of the PRAES to predict writing achievement was also demonstrated. However, data was collected which does not necessarily corroborate the prediction of the typology of students based the PRAES. This may be due to the nature of the measures or perhaps because there is no actual predictive potential. As far as the sample is concerned, as well as being representative, relevant and of a broad spectrum (compared with the samples in other empirical studies), it also allows us to describe the

other significant variables, which we have therefore excluded from the model.

students' achievement according to the type of practice employed by teachers.

composition, based on the differences according to the students' typology.

As regards the instruments used, given the revision of empirical and theoretical studies published in recent years, we can confirm that the PRAES and the applied writing measures (general and specific) used to evaluate both general achievement and specific aspects of written composition display not only acceptable validity and reliability but, also, appropriate sensitivity to the detection of differences based on the type of student and according to the prediction between variables. It is important to highlight that we know of no published studies that jointly employ the four PRAES components (Opinion, Approach, Classroom Behavior, Self-efficacy) to evaluate teachers, as well as the instruments applied to evaluate and to measure students' general and specific achievement in written composition, and that link both teacher and student measures. This justifies and affords relevance to the present study. As concerns the statistical analysis and its contribution to the study, when taking the intersubject factors – the three typologies of students (ADHD, LD, without LD) – as dependent variables, the different measures obtained for students and teachers verify the differences among the teachers regarding the role of their practice in teaching written

The results obtained are of high statistical significance in most of the dependent variables regarding students and teachers, with large effect sizes, in general. As regards the students, when belonging to one group type is taken as a grouping variable (with ADHD, with and

allow us to observe variables that are not always of the same nature.

**4. Discussion and conclusions** 

#### **3.1 Multiple linear regression analysis**

#### **Prediction of student typology**

If we considering the total writing measures (product totals from reader and text-based measures) and the measures of the PRAES (scale and the subscale totals) as predicting variables, and the type of student as the predicted variable, we obtain statistically significant results. When the measures are taken only from writing, statistically significant results in the prediction of the type of students were also obtained. However, when only the PRAES results are considered, they do produce statistically significant results in the prediction of the type of student. In any case, this is interesting, as it does not support the idea of there existing differences in the teachers' practices based on student typology. When the type of student is taken as a dependent variable (ADHD, LD, without LD) or as a variable predicted by the set of the total measures of the written product (students' achievement), in the hierarchic multiple regression step-by-step analysis, we obtained one model with a statistically significant regression coefficient (R² corrected =.496). The variables in the model that reached statistical significance are, among others: writing, reader based evaluation concerning order and structure [β = .711; t = 7.557; p = .001] and other aspects of total coherence [β = -.283; t = -2.447; p = .018]; those from the PRAES as regards classroom practice and the procedure subcomponent [β = -.237; t = -2.645; p = .011] and for the opinion and the motivation subcomponent [β = .182; t = 2.167; p =.035]. The remaining variables were excluded from the model, as they did not reach statistical significance.

#### **Prediction of the written product**

When the writing variables (the product totals from the reader and text-based measures) are taken as predicted variables and the PRAES variables as predicting, many produce statistically significant results, albeit of a low level. However, this is interesting, as they indicate a tendency, and given the nature of the different measures this points towards results which are very relevant from a theoretical point of view.

The fact that some total measures from the PRAES predict some totals for the writing product is very interesting, given that there is some relationship between the two. For example, the factors that are involved in the prediction of total productivity include the factors attributed to the family and the formal teaching approach, as are those used by teachers who will mainly employ these in students without LD. Other variables regarding teaching staff that predict some measure of achievement in writing are the procedures used (in several variables), the role assigned to the family, the materials used (in several variables), the natural teaching approach (more used with the group of students without LD and those with ADHD), personal self-efficacy, the formal teaching approach, teacher training, motivation, classroom practices, the use of plans and revision.

Of the 27 regression analyses extracted to predict writing based on PRAES, 23 were found to have statistical significance. The variables from the PRAES that are predictive regarding writing concentrate on the family subcomponent, classroom activities with the subcomponents of procedures and materials, in addition to the formal theoretical approach, the natural learning approach, personal self-efficacy, teacher training, motivation and the aspects of planning and revision. In addition, the attempt to predict writing achievement based on total productivity and the number of words parameter from the PRAES was carried out and a corrected R² of .067 was obtained. This is significant, although it only generates a regression model or equation with the predicting variables of PRAES for opinion with the family subcomponent [β = .276; t = 2.817; p = .006] and for approach, with the subcomponent of formal approach [β = .217; t = 2.213; p = .029]. It does not generate any other significant variables, which we have therefore excluded from the model.

In summary, regression analysis provides statistically significant data for predicting all the writing variables within PRAES, which presents interesting data regarding predictive validity. It also indicates the great predictive potential of the instruments applied, as they allow us to observe variables that are not always of the same nature.
