2.1 Participants

Eighty-eight undergraduate students participated in the research, divided into teaching courses of Physics, Chemistry and Mathematics. The participants were selected through applications to participate in the course offered. As criteria for the selection, it was considered: teaching undergraduate students of one of the three areas of knowledge and being enrolled in the last year of the course. Table 1 shows the characterization of the participants.

Regarding the location of participants, the highest concentration was in the state of São Paulo with 39, followed by Bahia (10), Minas Gerais (9), Mato Grosso do Sul (6), Pará (6), Rio de Janeiro (6 ), Espírito Santo (5), Paraná (2), Amazonas (1), Piauí (1), Goiás (1), Rio Grande do Sul (1) and Santa Catarina (1).

2.2 Data collection instruments

A teacher education program was the resource used to collect data in order to achieve the purposes of the study. For that, the following instruments used were:

Participant characterization form: Responsible for collecting information relevant to the characterization of the participants' profile. It was an instrument developed by the researchers and made available to participants via Google Forms.

Likert Scale of Social Attitudes Towards Inclusion (ELASI): it was an instrument for measuring social attitudes towards inclusion (^{Omote, 2005}). It was elaborated by the research group Difference, Deviation and Stigma, of the Paulista State University "Júlio de Mesquita Filho" (UNESP), Campus Marília. ELASI is organized by two equivalent forms (Form A and Form B), each consisting of 35 items, 30 of which are responsible for measuring attitudes, and the other five make up the lie scale that acts as an indicator of reliability of the respondents' answers, as their answers are completely predictable. Among the 30 items responsible for measuring social attitudes, 15 are positive items, which indicate agreement with the statement; and the other 15 are negative items, indicating unfavorable attitudes towards inclusion. Each of the items is made up of a particular statement for which one of the answers is expected: entirely agree, partially agree, neither agree nor disagree, partially disagree and entirely disagree. ELASI was the instrument responsible for the measurement in this study and was sent to participants via Google Forms.

Questionnaire of Course Social Validation: it was a script of questions adapted from studies already developed in the research group and was responsible for allowing an evaluation of the training process offered to participants from their perspectives. It consists of two parts, the first with 19 topics to be evaluated whose responses were Likert type with five possibilities of answers ranging from Unsatisfactory to Very satisfactory. And the second part consists of open questions for collecting feedback from the students.

2.3 Intervention

To carry out the expected intervention, a full distance teacher training program was designed by Moodle, called Inclusive Exact Science Teaching. The program had a total workload of 30 hours and was composed of three modules, each with a different theme: History, Policies and Pedagogical Practices. Each of the topics/modules of the course contained three texts or a study guide and three activities, being a discussion forum with a guiding question/situation accompanied by a problematization that guided the discussions, as well as two other activities to be carried out according to previous guidelines defined by the researcher/teacher. All modules were structured with materials (texts and activities) equivalent in quantity and level of difficulty, so that none stood out or interfered with the measurement performed.

Participants were divided into three groups and each enrolled in a room in which the course took place. The distribution of students in the classrooms occurred by random drawing. The objectives of the training program were:

Lead the student to understand, debate and be aware of the history of the exclusion of differences in school.

Provide the student with knowledge of the legal foundations of Special Education from the perspective of school inclusion in Brazil.

Give the opportunity to develop skills and knowledge of strategies that favor participation and learning from an inclusive perspective.

The aim of the research was to verify if, in particular, any of the three themes mentioned above would be responsible for promoting greater change in social attitudes in the students. Thus, to accomplish these purposes, the experimental arrangement comprised the construction of three rooms in Moodle, all identical in content. However, they presented alternation between modules, so that each group of students could start the course by a different theme. Thus, it was possible to measure attitudes separately by theme. Table 2 shows the adopted dynamics.

Each tab of the course, as well as all content contained therein, was made available for access for certain periods and closed for the next tab opening. Students had free access throughout and participated according to their own organization. All discussion forums were mediated to raise and maintain discussions among participants. After the data collection period, participants were given access to other themes.

2.4 Data treatment

The first stage of ELASI's quantitative analysis process included the assignment of numerical values to participants' responses to scale alternatives. According to the alternative selected by the participant, values were assigned: for positive items, 5 for the alternative entirely agree; 4 for alternative partially agree; 3 for alternative neither agree nor disagree; 2 for partially disagree; and 1 to entirely disagree. For negative items, the sense of value assignment was reversed. Excluding the lie scale scores and summing the values obtained in the positive and negative items, it was possible to obtain the total scores for each participant, which may vary in the ranges from 30 to 150. The lie scale received a specific score, being 0 for the expected answers and the value 1 otherwise. Thus, the closer to 0 the value obtained, the greater the reliability of the data for the measurement. Microsoft Office Excel programs and IBM SPSS Statistics 21 software were used for data processing and analysis, respectively. For comparative analyzes of the effect of the intervention, the so-called hypothesis tests were used (^{Dancey & Reidy, 2006}) and the statistically significant effect was considered when p <0.05.

For all data sets, the Shapiro-Wilk normality test was initially performed, in which the result directed the use of the appropriate test. For intra-room comparative analyzes, the nonparametric Wilcoxon test was used; and for the analysis between rooms, the Analysis of Variance (ANOVA), because the p value indicated that the data set presented the normal behavior. The analysis of variance made it possible to perform comparative analyzes of the performances of the three rooms and to reach the conclusion that was sought, that is, if any of the three themes tested had greater potential to promote changes in attitudes. The results obtained are presented below.

3.1 Intra-room performance analysis

Initially, we present the results regarding the intra-room performance analysis. The lie scale scores for the three rooms ranged from 0 to 1. Thus, this result indicated reliability in the participants' responses, since "the lower the final score, the greater the reliability of the answers given by the participant, ensuring that the task was performed with care and reliability" (^{Baleotti, 2006}, p. 64).

Table 3 presents the descriptive results regarding the pretest and posttest participants' scores for each of the three rooms, with the last column referring to the test result comparing the pretest and posttest measures for each of the rooms.

Considering that the score of each participant could vary between 30 and 150, it can be considered that all participants already started the intervention with high scores, since it was observed that the minimum values measured were already close to the maximum possible score value for ELASI. This data indicated that the participants already had favorable social attitudes towards inclusion even before starting the training program.

A quick examination of the data in Table 2 indicated fluctuations between the score values, considering both the pretest and posttest measurements for both rooms, both positive and negative. In room 1, for example, it was noted that the minimum value of the posttest was higher than the minimum value measured in the pretest. For the posttest of this same room, a decrease of one score unit was noted.

Regarding the effect of the isolated themes (1st module of the course), although it could be noted small fluctuations in the scores measured in the posttest compared to the pretest, it was found that none of the three rooms showed statistically significant changes in the attitudes of the students, given that in all the p values remained greater than 0.05, considering 95% of significance. This result showed that there was no significant difference between the pretest and posttest measures for any of the rooms.

Analyzing the participants' scores individually (Figure 1), three situations were present: i) score values that were higher in the posttest compared to the pretest; ii) cases of decreased scores; and iii) cases in which there was no change in the value measured between the two tests.

Source: The authors based on research data.

Figure 1 List of the students of the three rooms with their respective scores measured in the pretest and posttest. 

Room 1 was the one with the highest number of dropouts of participants right in the first phase of the course, which comprised the period of data collection. However, it was noted that this group of participants concentrated a greater amount of increase in scores than decreases in the posttest compared to the pretest, as can be seen in Figure 1; thus, there was a higher amount of higher posttest score compared to the pretest for this room. It is worth noting that, for the most part, by qualitatively analyzing the performance of these participants, it was noted that they were people with good dedication to the course. In the total of 22 students, six had lower posttest scores compared to the pretest (decrease) and two had the same measure in both tests.

Among the students who showed a decrease in attitudes, the participant identified by number 2, who only performed one activity of the unit, corresponded to a discussion forum in which the student inserted only one post. Thus, their insufficient participation could be a justification for the considerable decrease in scores.

Regarding the two other students who did not show changes in the score values (identified by numbers 1 and 20), it was noted that both already showed values close to the upper limit of ELASI. It is also worth mentioning that, among them, participant 1 performed well in the course, with interaction and attendance in the discussion forums and quality in the activities delivered.

In Figure 1, it was also possible to follow the distribution of the 33 participants who were enrolled in room 2 of the course, represented by their scores measured according to their identification codes. Of the 33 participants, 12 of them had a decrease in score, with the most intense cases of participant 31, with a decrease of 15 points, followed by participants 33 and 39, which presented a decrease of 13 points.

In addition to these cases of higher values of decrease, four course participants were identified who did not show changes in both measures, remaining with the same posttest score. It should be considered that such students already had scores very close to or higher than the average score for this room. Inserted in these cases, participant 49 stands out, who had an excellent performance throughout the course, always presenting reflection on the activities, punctuality and amount of participation. There are also cases of students who presented a considerable increase in score, highlighting participant 43 who had excellent interaction in the discussion forums, presenting attendance and content in the posts.

Regarding room 3, when analyzing the individual performance of the students, it was noted that 14 of them had a decrease in the values of measured scores, comparing the posttest with the pretest, consisting of almost half of the participants in the room. Among these decreases, we highlight participant 88 who presented a considerable decrease of 13 points. This was a student who presented participation in only one activity (a discussion forum) and dropped out of the course after the first unit.

3.2 Performance analysis between rooms

Although statistical tests did not report significant intra-room difference, fluctuations in score values between rooms were noticeable. Thus, we questioned whether such fluctuations would represent significant differences in order to assert that some theme had considerable influence on the measure of participants' scores compared to another. Thus, we present next the results elaborated based on the restlessness: Did the thematic variable influence the specific performance of any room?

For these analyzes, gross scores were no longer considered, but rather the performance of each participant by their respective room. In this way, the variable went from two measures (pretest and posttest) by the difference between them, in order to provide the performance of each participant. Under this data set, normality was tested by the Shapiro-Wilk test, reporting a p value of 0.260, therefore >0.05. The result indicated the normality behavior of the data and, thus, we proceed with the Analysis of Variance (ANOVA), whose result can be followed in Table 4.

The result (F = 0.215; p = 0.807) of this analysis indicated that the thematic variable did not influence the participants' performance in relation to the ELASI score, that is, there was no statistically significant difference between the performance of the three rooms. It is emphasized the need to consider that the initial values of pretest scores, that is, all were already close to the maximum possible for ELASI, so the research sample already showed indications of attitudes favorable to school inclusion.

The result presented in Table 4 made it possible to understand that, considering the context in which the research was developed (participants, strategy used, course content, etc.), the three themes tested were equivalent to each other regarding change to social attitudes. Considering that the themes tested in this study are common to be present in teacher training programs, it can be said that we are on the right track on "what to teach?". In ^{Souza's research (2013)}, the themes tested are among those present in the undergraduate curricula investigated by the author, which indicated exactly this trend of the path taken to train teachers for Special Education, taking as reference "principles, concepts and conceptions of the field of special education; education policy and legal guidelines; characterization of disabilities; teacher training and practice; specialized educational service and care programs" (^{Souza, 2013}, p. 138).

However, the issue is that, even if these themes are equivalent in the measures of social attitudes, there is still the problem of teachers who feel unprepared to work with TPSE, in addition to many classes in regular education without the adoption of inclusive practices by part of the teachers.

The results of this research corroborate the findings of ^{Woodcock, Hemmings and Kay (2012)} who, through a five-month inclusive education discipline, sought to explore the concerns and beliefs of teachers in training and how they could change through exposure to practical experiences and content related to inclusive education. The authors concluded that the results of the measurements before and after the intervention indicated that they were not significant in changing the participants' scores.

Given the duration of the training program used as an intervention, a total of 30 hours, it could be inferred that it was insufficient to bring about significant change, although it would be complex to determine what would be a more complete and considerable duration. The study by ^{Woodcock, Hemmings and Kay (2012)}, for example, had a longer course than offered in this research; however, it showed similar results.

The findings of ^{Wilkins and Nietfeld (2004)} also corroborate the results of this research. The authors agree that "attitudes towards inclusion are probably quite stable beliefs that are built on numerous previous experiences rather than driven by situational factors" (p. 119) - although, among participants, most already had favorable attitudes with values close to the maximum of the scale.

^{Tait and Purdie (2000)} also presented a study with the same purpose and with a considerable sample of 1,626 students from a university in Australia. In the study, the participants' social attitudes change was measured, whose intervention consisted of a one-year training course. According to the authors, the scores measured at the beginning of the course, when compared to those at the end of the course, were "unsuccessful in inducing desirable attitude changes" (p. 36). The authors questioned whether a longer course, for example four years, would be more successful in promoting these changes and listed "the need for researchers to assess the desirability of building into teacher training courses greater opportunity for contact between student teachers and people with disabilities" (p. 37).

With similar results, ^{Hastings, Hewes, Lock and Witting (1996)} indicated that they did not achieve attitudinal changes because the course used as an intervention did not offer participants contact with students with disabilities. In this sense, in ^{Santana's (2013)} research, among the groups of study participants, there was one in which was composed of university students whose classrooms had students with disabilities enrolled. The results showed that, in these classrooms, where students with disabilities were enrolled, the peers showed social attitudes more favorable to inclusion. Would this be an alternative for teacher training courses: to make it possible for future teachers to live with students with disabilities?

3.3 Final course evaluation

The complete training program was concluded with 49 students, considering at least 75% participation in the activities. There were 16 students from class 1, 19 from class 2 and 14 from class 3. Considering the dropout rate of distance courses offered (about 50%), it was considered that there was a dropout consistent with the expected.

Graduating participants were asked to fill in a questionnaire to collect their feedback regarding the course. The following items were evaluated: i) course structure (virtual environment, duration, workload, workload and duration relationship); ii) class structure (order of activities and themes); iii) supporting material (reading time and quantity relationship, quality of information covered and clarity of content); iv) trainer (clarity in subjects, didactics and mastery in the subject); v) participation as a student (activities and study of the material); vi) relationship of the course with future teaching practice (use of activities for reflection and preparation for practice with TPSE); and vii) satisfaction with the themes and activities (topics, discussion forum and assignment).

Quantitative analysis of these items varied on a Likert scale from Very Satisfactory to Unsatisfactory. Overall, for all items, participants kept their ratings on the Very Satisfactory and Satisfactory items. They were critical when evaluating their participation as students and some reports indicated this dissatisfaction:

[...] I think I could have paid more attention to the course, but this semester was very busy. (Student 46, research data).

[...] I feel sorry I couldn't dedicate myself as much as I would like. (Student 54, research data).

At the end of the semester it was a little difficult to dedicate myself to the course the way I wanted. (Student 58, research data).

Regarding the evaluation of the three themes approached, it was noted a good reception of all, as in the feedback of the student 3 when referring to the theme of pedagogical practices: The practice. As, for me, it was the last theme to be studied, I had already studied the two previous ones, the Practice happened more clearly (Student 3, research data).

It was noticed that, individually, the themes could fulfill their goals set in the planning, as the discussion about the laws: "The legislation allowed us to understand which way we are going" (Student 45, research data). This was a purpose for the discussion of this theme, to show the students the path from the legal point of view that the education of TSPE students has been following.

The theme of history for which it was intended to show the historical path of exclusion until sensitizing to guarantee the right of all was corroborated by the speech of the students:

The topic on which the evolution of social perceptions about people with disabilities was worked, as I believe that history can help us understand the present day, and information about it is essential to propose actions to combat prejudice and intolerance (barriers for the maintenance of an inclusive society). (Student 47, research data).

In general, it can be verified the potentiality of these themes for the development and maintenance of training programs.