| Date | November 2019 | Marks available | 9 | Reference code | 19N.Paper 3.HL.TZ0.3 |
| Level | HL only | Paper | Paper 3 | Time zone | TZ0 |
| Command term | Discuss | Question number | 3 | Adapted from | N/A |
Question
The stimulus material below describes a study on the influence of knowledge of stereotype threat on women’s math performance. Stereotype threat means that people believe a negative stereotype about themselves.
Previous research on stereotype threat and math suggests that women who are reminded of their gender before taking a math test will underperform compared to women who are not reminded of their gender.
The aim of this study was to test if teaching about the potential effect of stereotype threat before a math multiple-choice test helps women to perform better.
A convenience sample of 80 female university students taking a course of introductory statistics (N=80) and with a mean age of 19.5 years was selected. Participants received extra credit for participation. The female experimenter informed participants about the study and before the participants signed an informed consent form, she informed them about their rights to confidentiality and anonymity and about their rights to withdraw themselves or their data at any time. They were not fully informed about the purpose of the study until debriefing.
The participants were randomly allocated to one of two conditions:
- Condition 1 (math-test): participants were told that they would take an easy standardized math test for a study on gender and mathematical performance.
- Condition 2 (math-test and teaching intervention): participants received the same instructions as in condition 1, but the researcher also gave a brief lecture on the stereotype threat and said that women could experience anxiety due to the negative stereotypes of women and math. However, a stereotype had nothing to do with them and how well they might do on the test.
All participants were asked to write their gender on the tests, and then they had 30 minutes to complete the math test.
The results showed that participants in condition 1 scored lower than participants in condition 2.
The researcher concluded that knowledge of the stereotype threat had resulted in the better performance in condition 2. They suggested that teaching about stereotype threat could help other women to attribute anxiety about math to the stereotype and not to themselves.
Discuss how the researcher in the study could avoid bias.
Markscheme
Refer to the paper 3 markbands when awarding marks. These can be found under the “Your tests” tab > supplemental materials.
The command term “discuss” requires candidates to offer a considered and balanced review of how a researcher could avoid bias.
Biases in research may originate from design of the experiment, the researchers, as well as the participants.
Possible ways for the researcher to avoid bias in this study could include but are not limited to:
- Researchers could reduce bias by having a well-designed research protocol that explicitly outlines how data is collected and analysed in this experiment.
- The researcher could conduct a pilot study in order to test the suitability of the overall design, procedures and measures used in the experiment (for example, with regard to operationalization of variables) to see if a cause–effect relationship can be established between the IV and the DV (internal validity) .This would also help to see if all possible confounding variables have been taken into account. However, a pilot study may not be possible due to time restraints or lack of resources.
- A pilot study is an important step in ensuring construct validity, that is, making sure that the study in question actually is measuring ‘stereotype threat’ in relation to math so that the results can be generalized and used for prediction.
- Sampling bias (selection bias) is a danger in the case of a non-probability sample, as in this study. Although sampling bias may occur when participants in a sample are not selected randomly, but participants can then be randomly allocated to the experimental conditions in order to control for participant bias. This was also done in this study. Random allocation may increase the possibility of generalization. Another way to avoid sampling bias is to have a random sample but this is often not done in research like this one with a student sample.
- To prevent experimenter bias (researcher bias, the Rosenthal effect), the researcher could ensure that the experimenter is blind to the hypothesis of the study. This would help prevent threats to external validity. The researcher should also be aware of personal biases when formulating a research question and analysing data.
- The researcher can control for demand characteristics (i.e. participants respond to cues in the experiment, which somehow tell them what is expected of them) or the Hawthorne effect (i.e. the mere fact of being in a study makes participants perform better). This could affect their behaviour in this experiment and thus affect internal validity of the study. A possible way to control for this is using some degree of deception, which was also the case in this experiment.
- The researcher could control for bias related to having a male experimenter in a study with only female participants by having a female experimenter conduct the experiment. This was also the case in this study.
- The researcher could try to avoid confirmation bias and gender bias during analysis of data by having other researchers participate in the collection, analysis, and interpretation of data (researcher triangulation). This is important with regard to generalization of results, especially in a study with a single sex sample and a sensitive topic related to stereotyping.
Arguments based on a conceptual framework related to qualitative research, for example, personal reflexivity should not be credited.
Marks should be awarded according to the descriptors in the markbands. Each level of the markband corresponds to a range of marks to differentiate candidates' performance. A best-fit approach is used to ascertain which particular mark to use from the possible range for each level descriptor.
