Social change, cultural evolution, weaving apprenticeship, and development: informal education across three generations and 42 years in a Maya community

Abstract

Analyzing three sets of video data collected in one Maya community, we examined apprenticeship and learning of backstrap loom weaving over three generations spanning the years 1970 to 2012. Like many cultural groups, the Maya of Chiapas are experiencing rapid sociodemographic shifts. Three generations of girls (N = 134) were observed at their looms: in the 1970 subsistence economy; in the transition to a commercial economy in the 1990s; and in 2012, when the commercial economy required formal education. Multilevel models showed that intergenerational sociodemographic change - increased time in school, greater involvement in the money economy, and decreased family size - changed weaving apprenticeship, which, in turn, was related to changes in characteristics of learners. In 2012, weaving learners received more explanations, praise, and body instruction from their teachers. Learners, in turn, asked more questions. However, these changes came at a cost - the gradual loss of weaving as an everyday subsistence practice and art form. Tracing intergenerational change over three generations, this study makes a unique contribution to an understanding of cultural evolution.

Data availability statement

The data that support the findings of this study are available from the corresponding author, Ashley Maynard, upon reasonable request.

Results

We begin with the macro level of distal learning environment and the microlevel of proximal learning environment. We see each as each composed of a suite of interacting and synergistic variables.

Hypothesis 1: Ecology: Sociodemographic components of the distal (macro) learning environment will show positive intercorrelations among family involvement in commercial activities, use of mobile technologies, learner’s schooling, and mother’s schooling, as well as negative correlations with family size

Table 2 shows the correlations among the standardized sociodemographic variables used to construct the composite sociodemographic distal learning variable. This table shows that these variables are moderately intercorrelated. Most of the correlations are statistically significant, but each variable also explains unique variance. These correlations are in line with the theory of social change, culture, and human development (Greenfield, 2009); it posits that, at each level of the theory, variables operate synergistically rather than independently.

Table 2. The distal learning environment: Correlations among sociodemographic characteristics - number of children, commerce, mobile technology, own schooling, and mother’s schooling.

Hypothesis 2: Components of the proximal (micro) learning environment will show positive intercorrelations among body instruction, teacher explanations, and teacher praise

As with the sociodemographic variables, the proximal learning environment variables are, as predicted, moderately intercorrelated (Table 3). All correlations are statistically significant, but each variable also explains unique variance. Once again, these correlations are also in line with the theory of social change, culture, and human development (Greenfield, 2009) that, at each level of the theory, variables operate synergistically rather than independently.

Table 3. The proximal learning environment: Correlations among body instruction, teacher’s explanations, and teacher’s statements of praise.

Sociodemographic changes across the three generations: the distal learning environment

First, we were interested in demonstrating how the distal learning environment had changed over the three generations, so we examined engagement with schooling, involvement in commerce and textile commerce, and family size. Because the details of this historical process are crucial to the theoretical framework and the statistical models being tested, we include these shifts as part of the results, rather than as “background” factors placed in the Participants subsection of Method.

Hypothesis 3: Girls’ schooling will increase

We were interested in the endpoint of schooling for most girls. Because post-primary education was very rare, we analyzed the schooling of girls age 13 years and up. We found that girls’ schooling went up significantly over the three generations: Using just participants age 13 and up (F (2, 55) 25.82, p < .001, partial eta squared = .48, large effect size). A Games-Howell post-hoc test for multiple comparisons indicated that the difference between Generation 1 and Generation 2 was significant at less than .05, whereas the difference between Generation 2 and Generation 3 was significant at less than .001. Controlling for age as a covariate, schooling still went up significantly across the generations (F (3, 54) = 30.76, p < .001, partial eta squared = .53, large effect size). The mean years of schooling for girls ages 13 and up were: Generation 1 (0), Generation 2 (.60), and Generation 3 (4.76). No girls 13 and up had ever been to school in 1970. In 1991 and 1993, the range for schooling was zero to six years. In 2012, the range was zero to 9.5 years. The increased range in Generation 2 indicated that some girls were completing elementary school, which had become functional for carrying out commercial activities. The still greater range in 2012 reflected the fact that that at least one girl had received some secondary education in a neighboring Zinacantec community, a possibility that had not been there one generation earlier.

Hypothesis 4: Mothers’ schooling will increase

There was a small increase in mothers’ schooling, which was statistically significant F (2, 131) = 3.51, p = .03, partial eta squared = .05, a small effect size). Mothers’ mean schooling in Generation 1 was 0, while it was .36 years in Generation 2, and .84 years in Generation 3. A Games- Howell test indicated that the significant increase occurred between Generation 1 and Generation 2 (p = .004). The increase between Generation 2 and Generation 3 was not statistically significant.

In 1970 (Generation 1), not one mother had any schooling. In the early 1990s (Generation 2), the range went from zero to three years of school. In 2012 (Generation 3) the range of mothers’ school experience went from zero to six years. Again, the range is more telling than the means because it signals increased educational opportunity. One can also see that, as schooling opportunity expanded, daughters in the youngest generation had much more opportunity than their mothers had had.

Hypothesis 5: Family commerce will increase across three generations

Family involvement in commerce went up over the three generations (F (2, 131) = 37.62, p < .001, partial eta squared = .37, a large effect size). Although the increase from Generation 1 to Generation 2 was the largest (Games-Howell, p < .001), involvement in family commerce continued to go up from Generation 2 to Generation 3 (Games-Howell, p = .02). Mean commercial participation in Generation 1 was 0, while it was 3 out of 15 items in Generation 2 and 4 out of 15 items in Generation 3. The range was zero in Generation 1, 0 to 4 in Generation 2, and zero to 4 in Generation 3. The fact that the range stayed constant while the mean increased in Generation 3 shows that the variety of commercial activities had remained constant, but more residents of Nabenchauk were engaging in them.

Hypothesis 6: Textile commerce, which had increased significantly from Generation 1 to Generation 2, would decrease in Generation 3

As expected, the overall ANOVA indicated a significant change across the generations (F (2, 130) = 8.89, p < .001, partial eta squared = .12). Our means for textile commerce show that, in 1970, the average weaving learner was not involved in any type of textile commerce, and the range was zero. In the early 1990s, the average weaving learner had participated in 4 out of 12 possible types of textile commerce; the range was 0 to 10. In 2012, the average weaving learner had participated in only two or three types of textile commerce; the range was zero to nine. The Games-Howell test indicated that the increases in textile commerce from 1970 to the early 1990s and 2012 were significant at the .001 level, as expected. While the expected decrease from the early 1990s to 2012 took place, it did not attain statistical significance.

Hypothesis 7: Family size will decrease across the generations

Family size decreased significantly over time (F (2, 130) = 10.944, p < .001, partial eta squared = .144, a large effect size. The mean number of children in the family was (to the nearest whole number): Generation 1: 7; Generation 2: 5; Generation 3: 4. The significant difference, according to the Games-Howell test, is between Generation 2 and Generation 3 (p = .001).

The range of the number of children in the family is also interesting. The ranges are as follows: Generation 1: 1-13; Generation 2: 1-10; Generation 3: 1-8. Range is probably more informative than mean because young families in each generation were likely to add children in the future.

Hypothesis 8. Infant and child mortality will decrease across the generations

A one-way analysis of variance showed that there was a significant decrease in infant and child mortality from a mean of almost 1 child per nuclear family in Generation 1 to a mean of about 1 child per 3 families in Generation 2 to a mean of about 1 child in 30 families in Generation 3 (F (2, 130) = 7.261, p = .001, partial eta squared = .100, medium effect size) . Using the Games-Howell Test, the only significant difference was the decline in infant and child mortality from Generation 2 to Generation 3 (p = .006). The range was from zero to five child deaths per family in Generation 1, zero to four child deaths per family in Generation 2, zero or one child death in Generation 3.

Hypothesis 9. Mobile technologies will be used in Generation 3

There were no mobile technologies in Generations 1 or 2. In sharp contrast, mobile technologies were prevalent in Generation 3. The mean for Generation 3 was between one and two out of four items. The range went from zero to four. A significant ANOVA (F (2, 125) = 37.370, p < .001, partial eta squared = .374, a large effect size) and a post hoc Games-Howell test showed that showed that the Generation 3 differed significantly from Generation 1 and Generation 2 (p <.001).

Historical shifts across three generations: Changes in the proximal learning environment

Hypothesis 10: Girls would require more instruction in body postures and movements in Generation 3, compared with Generations 1 and 2

During the first weaving cycle, we counted verbal and nonverbal body instructions as well as instructions that were both verbal and nonverbal. There was more total body instruction across the historical periods (verbal, nonverbal, and both together). F (2, 130) = 4.22, p = .017. partial eta squared = .060, medium effect size. Means are Generation 1 = .46, Generation 2 = .47, Generation 3 = 1.53. These means translate to about half the learners receiving no body instruction in Generation 1 or 2 and, on average, learners in Generation 3 each receiving two or three instructions concerning body posture or movement. The range went from 0 to 4 instances of body instruction in Generation 1, 0 to 6 instances in Generation 2, and 0 to 17 instances in Generation 3.

Body instruction and kneeling

We analyzed an example of the body instruction given to a weaver at age 9 years in 1991 and compared it with her daughter at the same age (See Figure 1 and its explanatory caption). In 1991, PP was able to kneel correctly for a long period of time, and her mother did not give her instruction in how to kneel properly because such instruction was not necessary. But in 2012, PP’s daughter E was unable to kneel, complaining that her legs hurt, and her grandmother, PP’s mother, gave her granddaughter a number of prompts to position herself properly. We attribute the inability to kneel to the fact that PP grew up kneeling around the fire to make tortillas and carry out other activities. Small chairs were preferentially allocated to the males in the family. But daughter E grew up with tables and chairs; she lacked experience kneeling. It is known that to maintain the ability to kneel requires regular experience in kneeling growing up (Bolger, 2010; Molleson, 2007/2016).

For the sample as a whole, all learners kneeled for the whole analyzed weaving segment in 1970. In both the 1990s and 2012, a few girls deviated and used other positions while weaving. However, this was still a minority pattern (4 out of 58 in the 1990s, 5 out of 62 in 2012). Hence, chi-square tests (Preacher, 2001) did not achieve statistical significance for generational change in weaving position.

Figure 1. (Top): This is PP at age 9, kneeling to weave in 1991. PP has never been to school, and she grew up kneeling, for example, while making tortillas. Note how she is rising on her knees with her feet straight out in back of her.

(Bottom): This is her daughter E in 2012, also age 9. E has been going to school since preschool and has not learned to weave. This is her grandma CM trying to teach her for our study. But notice the body position. At 5:38pm, she is kind of kneeling but with her feet splayed out. But it is too uncomfortable for her and 8 minutes later (at 5:46), she is sitting.

Hypothesis 11: Praise by weaving teachers would increase across the generations, as would the ratio of praise to criticism

Confirming this hypothesis, we found that weaving teachers gave more praise across the three historical periods (F (2,131) = 5.79, p = .004, partial eta squared = .081, medium effect size). More meaningful than means is the intergenerational increase in the number of weaving teachers who gave any praise at all. In Generation 1, not 1 weaving teacher out of 14 praised the learner during the first cycle of weaving. In Generation 2, 5 out of 58 teachers praised the learner. In Generation 3, 21 out of 62 teachers praised the learner during the first cycle of weaving. The Games-Howell indicated that the statistically significant difference was between Generation 2 and Generation 3. In Generation 2, the range was zero to three instances of praise; in Generation 3 the range was zero to eight. In addition, in accord with the hypothesis, the ratio of praise to criticism increased significantly across the three generations (Chi-square = 12.11, df = 2, p =.002).

Hypothesis 12: Explanations by weaving teachers will increase across the generations

Explanations by weaving teachers went from none in Generation 1 to 8 explanations provided by a sample of 58 weaving teachers in Generation 2 and 9 explanations provided by a sample of 62 weaving teachers in Generation 3. The range in Generation 2 was from one to four explanations; the range in Generation 3 was from one to three explanations. This shift from none to low frequency occurred between Generation 1 and Generation 2. Because inferential statistics are not designed to compare a difference between zero and low frequency, this generational shift did not attain statistical significance. However, the increase of a feature from none to some is an important development in cultural evolution because cultural features adaptive in particular environments are selected for and become more frequent over time (cf. Fog, 1999). In addition, as our later analysis will show, it was part of the synergy with other Gesellschaft variables and was part of a successful three-generational model that integrated all three levels of analysis, sociodemographic change (the distal learning environment), teacher change (the proximal learning environment), and learner change.

Learner change across three generations

Hypothesis 13: Girls would have less weaving expertise in the current generation

We tested this hypothesis by examining the number of items girls had woven prior to the lesson we videotaped. Although the average age of the girls was approximately the same across the samples in the three generations, girls in Generation 3 had woven the fewest items prior to the weaving lesson we videotaped; means were one prior item in Generation 3 (to the closest whole number) vs. two in Generation 1 and Generation 2. An ANOVA confirmed that there was a significant intergenerational difference (F (2, 131) = 5.19, p = .007. partial eta squared = .073, medium effect size); the Games-Howell test confirmed that the only significant difference was the decline between Generation 2 and Generation 3 (p = .004). The range went from zero to five, the maximum, in each generation. Because learners gradually learn to weave more difficult items as they get older, we also carried out an analysis of covariance with age as the covariate and generation once again as the independent variable. This analysis confirmed that there were significant differences in the number of items girls had woven in the three generations (F (2, 130) = 4.30, p = .016, partial eta squared = .062, medium effect size). As expected, age was a significant covariate (F (1, 130) = 93.84, p < .001, partial eta squared = .42, large effect size).

Hypothesis 14: Schooling would compete with weaving; in other words, girls would not be learning to weave because they were attending school

This hypothesis was confirmed: Controlling for age, weaving expertise was negatively correlated with schooling. Those with more schooling had a lower level of weaving expertise, partial r (128) = −.24, p = .006, two-tailed test.

Hypothesis 15: Question asking, a trait adaptive in school, would increase across the generations

The frequency of girls asking questions was not high, but we found that the percentage of girls asking questions increased from Generations 1 and 2 to Generation 3: Percentage of girls who asked the teacher at least one question in each generation: In Generation 1, 14% of the girls asked questions; in Generation 2, 12% asked questions, and in Generation 3, 37% of the girls asked at least one question. We found that generation made a significant difference in question-asking: F (2, 131) = 6.39, p = .002, partial eta squared = .089, a medium effect size. The range was zero to one in Generation 1, zero to two in Generation 2, and zero to three in Generation 3. A Games-Howell test indicated that weaving learners in Generation 3 asked significantly more questions than learners in Generation 1 (p = .005) or Generation 2 (p = .004). Generations 1 and 2 are not significantly different from each other.

Hypothesis 16: Integrated model of changes in the distal learning environment, the proximal learning environment, and learner behavior

The data on which these results are based allow for comprehensive tests of the larger theoretical claim of this research—that changes in a community’s sociodemographic characteristics will lead to corresponding adaptations of learning environments which will in turn foster different developmental pathways. We begin with the correlations.

As an indicator of individual development adapted to a Gesellschaft ecology, we used the learner variable of the number of questions a girl asked during the first cycle of her weaving. The higher values of generation, distal learning environment, and proximal learning environment represent a Gesellschaft orientation, and number of questions was positively correlated with these variables (Table 4).

Table 4. Correlations among generation, distal learning environment, proximal learning enviornment, and learner variables.

As an indicator of individual development adapted to a Gemeinschaft ecology, we used the learner variable of weaving expertise - the ability to weave increasingly difficult items. The lower values of generation, distal learning environment, and proximal learning environment represent a Gemeinschaft orientation, and weaving expertise was negatively correlated with these variables (Table 4).

In sum, we created two causal models that would explicate the factors producing historical changes in learner skills. One model (Figure 2) identified factors leading to the intergenerational increase in question-asking while learning to weave, a transformation of cultural transmission in the Gesellschaft direction. The other model (Figure 3) identified factors leading to the intergenerational loss of weaving expertise, a key skill that was required and adaptive in the subsistence Gemeinschaft ecology of Nabenchauk in earlier times.

Figure 2. Test of the indirect effect of generation on the number of girls’ questions as mediated by the sequence of the effect of the distal learning environment on the number of girls’ questions through learning environment.

Note: Significant positive coefficients indicate that the younger the generation, the more the social ecology was Gesellschaft-adapted as indicated by the sociodemographic characteristics in the distal learning environment (a₁), and the more Gesellschaft-adapted the sociodemographic characteristics in the distal learning environment (b₁) and proximal learning environment (b₂), the more questions the girls asked. That the direct effect (c‘) is not significant means that the correlation between generation and girls’ questions is largely explained by the sociodemographic characteristics comprising the distal learning environment (a₁ b₁: CI = [.044, .641]) and the serial mediation of the distal learning environment through the proximal learning environment (a₁ d₂₁ b₂: CI = [.005, .295]).

⁺p < .068; *p < .05; ****p < .0001.

Figure 3. Test of the indirect effect of generation on weaving increasingly difficult items as mediated by the sequence of the effect of the distal learning environment on the ability to weave through the proximal learning environment.

Note: The significant positive coefficients indicate that the younger the generation, the more the social ecology was Gesellschaft-adapted, as indicated by the sociodemographic characteristics (a₁). The significant negative coefficient indicates that the more Gesellschaft-adapted the proximal learning environment, the fewer difficult pieces were woven by the girls (b₂). That the direct effect (c‘) is not significant means that the correlation in which younger generations were less likely to weave difficult pieces is largely explained by the serial mediation of the distal learning environment through the proximal learning environment (a₁ d₂₁ b₂: CI = [-.212, -.012]).

⁺p = .065; *p < .05; ****p < .0001.

Tests of serial mediation (Hayes, 2018) allow testing the predicted models that differences between the generations of weaving learners in the number of questions they asked in the learning session and differences in weaving expertise might be explained by differences in sociodemographic characteristics of the distal learning environment and features of the proximal learning environment during weaving. Importantly, in the predicted models sociodemographic characteristics of the distal learning environment followed by proximal learning environment should be the effective mediators between generation and girls’ questions.

The test of the whole model in Figure 2, with girls’ questions as the dependent variable and generation, distal learning environment, and proximal learning environment as predictor variables without considering specific paths of mediation, is significant (F(3,130) = 24.04, p < .0001.) Figure 2 shows the tests of the potential paths of mediation. As expected and illustrated in the figure, there was no direct effect (c‘) of generation on girls’ questions. Also as expected, generation does predict distal learning environment (a₁). When taking generation into account, both distal learning environment (b₁) and proximal learning environment (b₂) predict girls’ questions.

The tests for mediation entail combinations of these paths. In terms of the critical test of the total theoretical model, there is serial mediation (a₁ d₂₁ b₂: CI = [.001, .262]) such that there is an indirect effect of generation on girls’ questions through the serial mediation of distal learning environment followed by the proximal learning environment. These results confirm the predicted model in which each generation is proposed to have a different distal learning environment with distinct sociodemographic features; each distal learning environment then produces a different proximal learning environment - i.e., teacher-learner interaction while learning to weave. Finally, these differences in proximal learning environment then predict different pathways of children’s behavioral development as learners, which in this case is whether or not girls will tend to ask questions of the teacher during a weaving lesson.

The test of the whole model in Figure 3, with weaving expertise as the dependent variable and generation, distal learning environment, and proximal learning environment as predictor variables, without considering specific paths of mediation, is statistically significant (F (3,130) = 4.38, p < .01.) Figure 3 shows the tests of the potential paths of mediation. As with the previous model and as expected, generation does predict distal learning environment: participants from each subsequent generation are from smaller families, participate in more commercial activity; and receive more schooling; in Generation 3, participants also have mobile technology (a₁). When taking generation into account, a more Gemeinschaft-adapted distal learning environment (larger families, less commercial involvement, less school experience of both learner and mother, and less mobile technology) and a more Gemeinschaft adapted proximal learning environment, (less teacher explanation, less praise, and less body instruction during the weaving lesson) does predict greater weaving expertise (b₂).

The tests for mediation entail combinations of these paths. In terms of the critical test of the total theoretical model, there is serial mediation (a₁ d₂₁ b₂: CI = [-.201, –.003]) such that there is an indirect effect of generation on weaving expertise through the serial mediation of distal learning environment followed by proximal learning environment.

These results confirm the predicted models in which each generation is proposed to have a different distal learning environment with distinct sociodemographic features; the distal learning environment typical of a Gesellschaft ecology then produces processes of weaving apprenticeship - learning environment and child behavior - adapted to this ecology; but these Gesellschaft-influenced processes also lead to a decrement in weaving expertise.

Discussion

In this age of globalization, where cultural groups are experiencing rapid changes in economics, demographics, and daily routines, we aimed to study links between macro-level shifts and micro-level learning interactions. Using the Zinacantec Maya community of Nabenchauk in Chiapas, Mexico, as a natural laboratory, we found coordinated changes over a period of more than 60 years in the distal learning environment on the macro-level and the proximal learning environment and learner behavior on the micro-level of weaving apprenticeship. Commercial activity and education expanded for both learners and teachers, mobile technologies began to be used, and family size decreased—all features of the distal learning environment. As these ecological shifts occurred, the proximal features of intergenerational cultural transmission also changed: weaving teachers praised and explained more while criticizing less, and they responded to learners’ increased needs for bodily instruction by providing it. In turn, these historical shifts produced an increase in question-asking while learning to weave, a Gesellschaft-adapted skill valued in an educationally oriented society, and a decrease in weaving expertise, a Gemeinschaft-adapted skill valued in a subsistence-oriented community.

Fewer girls were learning to weave in 2012 because they were involved in other kinds of activities, including attending school, producing on order (for pay) specialized aspects of textile creation (e.g., drawing patterns for fancy embroidery), or working with their families in other kinds of commerce (Greenfield, 2004; Greenfield et al., 2009). Because many girls were not learning to weave, the number of items girls had woven prior to our videotaped weaving lesson had decreased significantly by 2012, controlling for learners’ ages. This was the case even though the average age of the girls was approximately the same across the three generations. The decline of weaving expertise is a major phenomenon of Maya cultural evolution; this is the case because backstrap loom weaving has been the most complex technology of Maya culture that has survived over the millennia.

Schooling competed with weaving. This finding demonstrates, on the level of individuals in a defined sample, the general anthropological conclusion that school attendance pulls children away from learning subsistence skills (Lancy, 2012). Even more interesting in terms of learning processes, girls who had more schooling behaved differently in the weaving lesson than those with less schooling. Specifically, weaving learners who had been to school asked more questions as they wove in front of the video camera.

Our integrated model showed that girls whose mothers had been to school also asked more questions. Asking questions is a form of self-assertion, a quality that is adaptive in a Gesellschaft ecology. Asking questions is, moreover, something that teachers do; and children are encouraged to do the same in school. Prior research has shown that going to school leads children to internalize the role of the teacher (Maynard, 2002), a role that includes question-asking. An important point is that the practice of asking questions in the context of formal education carried over and generalized to weaving apprenticeship, a process of informal education. In sum, as the evolution of community culture made backstrap-loom weaving less central, intergenerational transmission also evolved, reflecting cultural features that were gaining importance over time—most notably schooling.

The topic of socialization in developmental psychology does not generally include the socialization of the physical body and bodily movements. Yet weaving is both a bodily skill and a cognitive skill (Maynard et al., 1999; Greenfield et al., 2003; Maynard et al., 2015). Therefore, the use of the body was of great interest to us. Compared with Generations 1 and 2, girls in Generation 3 required more instruction in how to use their bodies in weaving. This is because kneeling, a long-established proper body position for Zinacantec Maya females, had diminished in favor of sitting in chairs, at home and at school (for those who went to school). We believe that the everyday use of chairs and the decrease in girls’ weaving experience are factors in the need for body instruction. Our case study of AE showed that body position could not be taken for granted by girls who went to school and had no experience in everyday life of kneeling to carry out various activities.

We have traced how the shift at the ecological level from Gemeinschaft to Gesellschaft engendered intergenerational changes in the learning environment, in turn shifting learner behavior in each generation. This cross-cohort comparative design has demonstrated how cultural evolution can be directly observed and assessed on multiple levels across multiple generations. Our findings provide an empirical demonstration of nested levels of environmental influence very much in the spirit of Bronfenbrenner’s (1977) Ecological Systems Theory.

But our research also documents that these multilevel changes come at a cultural cost - the gradual loss of weaving as an everyday subsistence practice and art form. In this too, the village is a microcosm of Mexico and the globalizing world.

Broader implications

These shifts in the macro-environment in the Gesellschaft direction are happening on a global level. Many societies and communities beyond our study site have moved in recent generations from agriculture and subsistence lifestyles to commerce, formal education, and technology (e.g., Abu Aleon et al., 2019; Weinstock et al., 2015; Zeng & Greenfield, 2015). In addition, migrants all over the world are moving from communities based on agricultural and subsistence lifestyles to host societies that have highly developed commercial, educational, and technological ecologies (e.g., Greenfield et al., 2019). Therefore, our longitudinal findings in Nabenchauk are applicable to shifts in learning environment and cognitive development brought about by endogenous social change and migration all over the world. The traditional view in developmental psychology that cultural differences are static does not take into account the fact that the influences on learning undergo massive change over time, making learning and cognitive development dynamic processes, and producing altered patterns of cognitive development (Maynard et al., 2015; Weinstock, 2015).

As Zinacantec girls over time asked more questions as they and their mothers were exposed to more formal education, their behavior became more like the children teachers expect to teach in the U.S. or Western Europe. This finding illustrates another broader implication of the present research: the global direction of social change in the Gesellschaft direction often has the effect of reducing cultural differences in child development (e.g., Zhou et al., 2018).

However, lest we mistakenly view the emerging patterns of development as progress, we need to remember the cultural losses from movement away from a Gemeinschaft ecology. On the physical level, we have documented loss of the ability to kneel as the human-made environment has changed. In Chiapas, Mexico, weaving as a cultural practice is a loss in Maya communities across the state resulting from the increase in educational opportunity (Zambrano, 2000). Such losses are also applicable to communities in other countries moving from subsistence life styles to life in Gesellschaft environments. For example, with immigration to Israel, Ethiopian immigrants now value formal education and technology, but devalue their artistic practice of clay sculpture, originally a subsistence practice in their ancestral country of Ethiopia (Greenfield et al., 2019). While we need to understand that there are developmental losses as well as gains from ecological movement in the Gesellschaft direction, we also need to remember that each set of cultural practices is adapted to a different set of ecological circumstances.

Limitations

It would be interesting to know if delayed childbearing was a factor in producing smaller families. Because we have not calculated each mother’s age when she bore her first child, we do not know if this was a factor in family size. We will be able to assess this factor in the future. Undoubtedly, some mothers continued to have children after the study; and these could not be counted in our assessment of family size. This is a limitation on our accuracy in reporting family size, but it is a constant across the generations, so does not affect our statistical comparison across the three generations.

Another limitation of the present study is that the coders of the videos were not blind to all hypotheses, particularly because we had published articles together based on the first and second generations. However, the coding system was rather rigid (how many seconds of this, what grammatical structure of that sentence, positive or negative statements). It would be hard to alter the coding systematically to fit a hypothesis. Furthermore, we achieved strong to perfect reliability on the coding of a random set of videos, which also adds to credibility. It would have been impossible to find coders who could weave, speak Tzotzil, and had enough formal education to understand the coding process in order to achieve blind coding.

Educational application and implications for cognitive development

Children’s questions have commonly been seen as a part of cognitive development that is essential to learning in formal and informal learning environments (Chouinard et al., 2007; Haber et al., 2021). Inquiry-based learning, which, as the name implies, centers on raising questions, specifically from learners, is often seen as the gold-standard of educational practice (Geier et al., 2008; Nicolopoulou et al., 2021). However, much educational practice and developmental research treats children’s question-asking and the types of questions they tend to ask as if it were a natural process rather than the product of adaptation to different learning environments. Many educators give a nod to sociocultural practices, seeing schooling and instructional methods as sociocultural practices, but still expecting that children naturally ask questions in the way they, the educators, do. Our research indicates that children’s questionasking is not natural but culturally specific, an adaptation to the learning environments and cognitive development of Gesellschaft social ecologies (Jegede & Olajide, 1995). The current research indeed found that the latest generation of weavers asked more questions of their teachers in the course of learning, which, we argue, is indicative of the change in the social ecology from the earlier generations toward more Gesellschaft-adapted learning environments and a subsequent shift toward Gesellschaft-adapted cognitive development.

These findings are very applicable for teachers. Many children in the United States and Europe come from immigrant families in which parents are from rural more Gemeinschaft environments than the environment of their host country. In such communities, questions can be considered a mark of disrespect; agricultural workers in California from Mexico or Central America provide an example of this cultural interpretation (Delgado-Gaitan, 1994). Children from such environments are socialized to show respect to parents and teachers by NOT asking questions.

What teachers can take away from this analysis and our findings is that children brought up in this way may not engage in the question-asking valued in school. When children have experienced this type of cultural socialization, teachers can learn from our research to treat the absence of children’s questions as a cultural trait rather than a cognitive deficit. However, because questions are valued in the host culture, teachers can also be explicit about the difference between home and school cultures and, early in their education, gently guide children toward the question-asking so important and valued in formal education.

In addition to environmental influences on the presence vs. absence of learner questions, there are also ecological factors in the types of questions that are relevant and valued. Much of the research regarding children’s questions, at least with U.S. samples, highlights different types of questions than the ones the most recent generation of weaving learners were asking. For instance, Chouinard et al. (2007) report that preschool children’s questions tend to be fact or causal explanation seeking. The Zinacantec weaving learners' questions concerned learning the procedure. We argue that the differences in the types of questions students ask in the weaving environment or in formal schooling are not just a matter of age difference, but also reflect the different goals of the two learning environments. So, children in school might not be asking about procedures, because theoretical knowledge and not procedures are the center of schooling; insofar as procedures are involved—such as in science class labs—these are not done for their own sake but in the service of theoretical learning. This contrasts with the tasks of teaching and learning to weave. But that different types of environment entail very different tasks is precisely the point. Teaching and learning, and children’s cognitive developmental tasks are adapted to the learning environment.

Direct application to classroom teaching is also possible. Author Childs, who teaches sixth graders in the U.S., has applied our research in a very concrete way to teach about cultural relativity, more specifically, to teach students that children are taught differently in different ecological circumstances. She sets up a backstrap loom in a classroom, demonstrates how weaving is done, and then asks a student to get into the loom. She then provides a demonstration of teaching as it was done in our first generation of weaving learners: criticism rather than praise; imperatives rather than questions; observation rather than doing; replication rather than innovation. She also explains that in such an environment, education takes place primarily at home rather than at school. She points out that these are teaching and learning procedures they have probably never even imagined exist. While many of the students find Childs' demonstration quite alien, a number of them come to understand something about profound cultural differences; hence the demonstration functions as a lesson in cultural relativity.

Childs also points out to her students that, as Zinacantec Maya environments have become more like ours, most recently with increased opportunities for formal education, teaching and learning procedures have also become more like ours. She uses her example of backstrap-loom weaving in Nabenchauk to teach that modes of teaching and learning are adapted to particular environments and that they therefore change as environments shift over time. This innovative application of the research is a concrete, valuable example of using it in an educational setting.

Acknowledgements

Our third wave of data collection (2012) was supported by grants from the Spencer Foundation, UC MEXUS, and the UCLA Latin American Institute. The UCLA Center for the Study of Women provided financial administration. Grateful thanks to our field assistant, Maruch Ch’entik, who also assisted with the second wave of data collection in 1991. Finally, we thank our study participants in Nabenchauk, many of whom are the third generation of their families to participate in our studies.

Disclosure statement

No potential conflict of interest was reported by the authors.