We interviewed students attending national scientific events. We went to three conferences in distinct regions: ecology (2013; Porto Seguro city, Bahia state), botany (2013; Belo Horizonte, Minas Gerais), and zoology (2014; Porto Alegre, Rio Grande do Sul). Despite this convenience sampling may entail some biases (i.e. including mostly the more dedicated, advanced, or financially secured students), we adopted the strategy in order to reach a larger and geographically diversified public.
Students’ evolutionary thinking was assessed through questionnaires. We adapted the procedure applied by the Gallup Organization, which has acquired USA citizens’ opinions on creationism/evolution since 1982 (Gallup 2017). Interviewees were asked to answer the question: Which of the following assertions better expresses your opinion about the origin and development of human beings?: (1) God (or another divine force) created human beings already similar to what they look like nowadays approximately 10,000 years ago; (2) Human beings developed over the millions of years from other forms of life, with God (or another divine force) guiding the process; (3) Human beings developed over the millions of years from other forms of life, but God (or another divine force) had no part in the process.
This survey may lead to biases and distortions of the actual views, starting with the focus on human evolution, which may lead those interviewed to feel uncomfortable (Kampourakis and McCain 2016). In addition, misconceptions may be related to the further classifications of the three answer options. The main caveat is perhaps option 2, which at the same time may assemble people prone to creationism and/or evolution. In fact, the most parsimonious conclusion is that answers related to option 1 establish a lower bound on the number of creationists in the sample while option 3 establish the opposite bound on the number of evolutionists. Even thought, we still used this methodology style because we intended to compare our results to previous studies on Brazilians common citizens and students (see “Discussion”). Here, we label options 1, 2 and 3 as “creationism” (Cr), “divinely guided evolution” (DGE) and “naturalist evolution” (NaE).
Since we were interested in the opinions of students currently enrolled in the biological sciences graduate course, we did not interview people who have completed their degrees (and stopped studying), students from other courses, or professionals. However, we also surveyed postgraduate students, but only those who graduated from the biological sciences course. We only considered stricto sensu (academic) postgraduates (students currently pursuing master’s or doctoral degrees). Concerning undergraduate students, we gathered information about their age, graduation period, educational institution, and area of interest (ecology, botany, zoology, or other area). Among postgraduate students, we collected data about their age, area of interest, and educational institution, including the name of the postgraduate program. Based on data of specific educational institutions, students were situated in one of the five geographic macroregions of Brazil: North, Northeast, Center-West, Southeast, and South. For further analysis, we relocated students into the three geo-economic regions: Amazonia, Northeast, and Center-South (sensu Geiger 1964). In the mentioned order, these regions present a crescent degree of richness and development (Geiger 1964). In order to avoid geographical biases, we interviewed only people currently studying in the same geo-economic region in which they were born.
For each interviewee, we attributed a score using numeric indicators of quality provided by the Brazilian government for each educational institution. We used indicators that not only referred to the students’ performance but also other elements of their educational environment, such as infrastructure and professors’ quality. For the graduate level, we took CPC (Preliminary Concept of Course) scores from 2011 (INEP 2016a). We used the “continuous values,” which range from zero to five. In the situation where institutions offer courses on two modalities (“Licenciatura” [teaching career] and “Bacharelado” [other assignments]), we extracted the average number. We did not attribute scores for courses that present “no concept” or for those that were not in the spreadsheet. Concerning postgraduates, we used the CAPES (Coordination for the Improvement of Higher Education Personnel) concept from 2013, in which values range from zero to seven (CAPES 2016). Since both the CPC and CAPES concepts are results from triennial evaluations, we selected those closest to the period in which the surveys were conducted.
The frequencies of opinions according to area (botany, ecology, and zoology), level (under and postgraduate), and macroregion (North, Northeast, Center-West, Southeast, and South) were compared in a multifactorial design by applying a log-linear model using the package MASS (Venables and Ripley 2002). Individuals who assigned “other area” (78 interviewees) in the area of interest were not analyzed regarding this factor. Besides the mentioned predictor variables, the whole model also comprised all possible interaction terms. After detecting effect for macroregions (see “Results”), we compared opinion frequencies according to geo-economic regions through a Chi squared test (5000 permutations simulated). Afterwards, we performed pairwise comparisons among the three regions, accepting a Bonferroni corrected P < 0.017 as significant to reduce type I error (Cabin and Mitchell 2000).
We investigated which factors may be related to individuals’ deviating opinions from NaE by submitting our continuous variables (age, graduation period, and institution score) to a mixed multinomial logistic regression using the package mlogit (Croissant 2012). In order to generate equivalence between graduate and postgraduate scores, we took the ratio value by dividing the obtained score by its maximum possible (graduate scores per five and postgraduate per seven). Students from institutions with no score were not analyzed regarding this factor. We checked for multicollinearity among the explanatory variables and considered it to be no problem in our data, as correlations among them were low (age vs. period: 0.37; age vs. score: − 0.04; period vs. score: − 0.03). Since effects of area, level, and macroregion were investigated previously with the log-linear model, these were treated as random effects. To test our specific hypothesis that divinely guided processes as a whole increase as institution quality decrease, we conducted a simple logistic regression grouping Cr and DGE.
In order to investigate possible institution score proportion differences, we fitted a two-way ANOVA with level and geo-economic region as fixed effects. Each institution was treated as a single replicate, regardless of the number of students belonging to it. Due to the unbalanced data for geo-economic region, we adjusted a “type II” sum of squares (Langsrud 2003) using the package car (Fox et al. 2016). Data distribution was examined by inspecting the homogeneity of the residual vs. fitted values plot (Zuur et al. 2009). Post hoc pairwise comparisons were performed using the Tukey’s HSD test. All analyses were performed within R statistical environment version 3.4.0 (R Core Team 2017).