Drs. Elizabeth Bonawitz, Patrick Shafto and I (Yue Yu) have studied the implications of parental consent on research findings. In 2017, we conducted a survey about the rate of parental consent in developmental studies. Results showed that substantial proportions of parents decline or ignore recruitment efforts from researchers, and the consent rate varied across different recruitment methods. Using empirical and modelling approaches, we further showed that when generalizing findings from lab samples to the general population, ignoring the role of consent may lead to biased estimation of means, as well as underestimation of standard deviations (Yu, Shafto, & Bonawitz, 2020). Here I describe the findings from the parental consent survey.
Sample
We circulated a brief survey among lab managers and project leaders through the email list of Cognitive Development Society, a leading society of researchers who conduct experimental research with infants and young children. We received responses from 32 researchers from 4 countries. The number of recruited participants from each response ranged from 6 to 2500 (median = 82), and all recruitment happened between 1990 and 2018 (with the majority from 2014 to 2017).
Consent rates
In our sample, 37.5% of researchers recruited children by directly contacting their parents through phone, email, or mail; 46.9% disseminated consent forms through children's schools, preschools, daycares, or after-school programs; and the remaining 15.6% recruited parents and children on-site, such as by directly recruiting and testing them in museums and camps. The average consent rate differed by recruitment method:
- Recruitment by phone, email, or mail: M = 48.1%, SD = 22.1%;
- Recruitment through school or preschool: M = 37.0%, SD = 24.7%;
- On-site recruitment: M = 84.4%, SD = 14.6%;
Factors that affect consent rates
Besides recruitment methods, we are also interested in examining how consent rates are affected by factors about the target population and the design of the study. The factors we measured include: the target age range of the child, the demographics of the target population, whether the study is one-time (vs. multiple sessions or longitudinal), whether the parent needs to be involved in addition to the child, the described length of the study, whether the study requires a trip to the lab, whether video or audio recording is required, and whether the study involves collecting physiological or neuroimaging data.
Results from a multiple linear regression showed that after controlling for recruitment method, two of these factors have significant effects on consent rate:
Results from a multiple linear regression showed that after controlling for recruitment method, two of these factors have significant effects on consent rate:
- Studies that only require children to participate have higher consent rates than studies that also require parents to be involved (β = .33, p < .05);
- Studies that recruit infants below 1 year of age have lower consent rates than studies that do not (β = -.44, p < .05).
Implications
The substantial percentages of non-enrollment may pose challenges to the generalization of findings from lab samples to the general population. By using a combination of observational, experimental, and computational methods, Yu, Shafto, & Bonawitz (2020) estimated that certain behavioral measures of children could differ significantly between those whose parents consented them to participate in research, and those whose parents did not. Moreover, the standard deviations are likely underestimated when generating from a sample of consented children to the general population. Our findings support and add to the recent concern about persistent sampling biases in developmental psychology (Nielsen, Haun, Kärtner, & Legare, 2017). In addition to the biases resulted from unrepresentative pools of participants researchers usually recruit from, the process of recruitment itself may also skew the sample. Yu, Shafto, & Bonawitz (2020) also provided a method that could be used to estimate and understand the biases related to parental consent, which could be used as an empirical base for recruiting a more representative sample.
Citing the survey results
To use the survey results in publications please cite:
Yu, Y., Shafto, P. & Bonawitz, E. (2020). Inconvenient samples: Modeling biases related to parental consent by coupling observational and experimental results. Open Mind: Discoveries in Cognitive Science, 4, 13–24.
Please note that some of the consent rates reported in this paper are slightly different from the numbers reported here. This is because we received a few more responses after the paper was submitted, and they were included here but not in the paper. The discrepancies are minor and should not affect the overall findings.