Womanhood represents the shared collective experiences of people who identify as women (Young, 1994) with respect for considering intersectional experiences that accompany race, sexuality, religion, among other identities and experiences (i.e., each individual’s experience is unique and there are shared elements of experience within social groups). Arendell (2000) presented a definition of motherhood that was not constrained to gender roles and instead described motherhood as occurring when an individual participated in the relational and logistical work of child rearing. Defined in this way, motherhood represents the collective experience of those who are functioning in the role of mothers (Knowles & Cole, 2014), again with respect for considering the intersectional experiences of mothers in a diverse society. Defining gender and parental roles as socially constructed and rooted in human language is central within feminist theories that emerged from within an American pragmatic tradition (Whipps & Lake, 2020).Footnote 1Although biological sex differences exist between people assigned male or female at birth, gender roles, norms, and expectations are arbitrarily applied (i.e., there is no justifiable biologically rooted reason that women should operate as the near exclusive primary caregivers in the society) as they evolved within patriarchal systems that imbue interpersonal and individual power to men over women along a binary and heteronormative view of gender and sexuality.

Differences in language around the gender binary is apparent in research spanning at least a half-century. According to research by Williams et al. (1975), adjectives most commonly associated with “female” and “womanhood” are drastically different than words most commonly associated with “man” or “manhood,” including descriptors such as affectionate, emotional, and fickle, that are differentially applied to women and not men. Sickman et al. (2023) applied an analysis rooted in Relational Density Theory (Belisle & Dixon, 2020) showing that relational framing patterns were evident along a binary view of gender and influenced the way participants perceived gendered scenarios. The potentially oppressive nature of these language relations around womanhood and manhood can be seen in many facets of experience and are evident within the gender pay gap. In 2020, women in the United States earned 84% of what men earned according to a Pew Research Center analysis of median hourly earnings for both full and part-time workers (Kochhar, 2023). Women, therefore, would have to work 42 additional days on average to make what their male counterpart made that same year (Kochhar, 2023).

These problems are also apparent within behavior analysis and must be solved internally as well as externally. Li et al. (2019) reported that salaries for male professors in the field exceeded salaries for female professors in the field at all ranks, ranging from 3% to 17% greater median salaries for male professors. Li and colleagues also conducted an analysis of participation of women in behavior analytic research as one metric of holding power positions within the field. The percentage of articles with women as first author ranged from 27.2% to 57.6%, which is problematic when over 80% of practicing behavior analysts identify as women (Li et al., 2018). Martin et al. (2022) also evaluated gender, racial, and ethnic disparities in research experience within the contextual behavioral science community, finding that men reported more publications and first authored publications than women, and white respondents reported publishing more frequently within the Journal of Contextual Behavioral Science than under-represented minority groups as compared to other journals. These data, combined with those reported by Li et al. (2018, 2019), suggest that power in the field may be predominantly held by men, mirroring gendered differences in the society more generally (Li et al., 2018).

Matricentric feminism centers the experience of women who are mothers as distinct in some ways from the experience of women more broadly given the social, political, economic, and psychological context operating around parenting and womanhood (O’Reilly, 2019). As noted by O’Reilly (2019), “mothers are oppressed under patriarchy as women and as mothers,” (p. 15); and, “Motherhood, it could be said, is the unfinished business of feminism” (p. 14). Motherhood can lead to interruptions in women’s career paths and impact long-term earnings (Kochhar, 2023) leading to systemic disadvantages, especially in the workforce (Kalev et al., 2018). A mother’s wage is negatively impacted by 4.4% after having a child, whereas a father’s wage is generally impacted less than 1% (MOTU, 2011). Being a father can even be beneficial for men in the workforce; instead of facing a penalty, fathers receive what is called the “fatherhood bonus,” where having children can result in an increase in wages due to fathers being seen as reliable, stable, and committed in all aspects of life (Budig & England, 2001). This role can also benefit fathers in times of work-family conflict (Luhr, 2020). During the coronavirus-19 pandemic, women carried a heavier load than men in providing unpaid work duties such as childcare, caring for the elderly, and higher housework responsibilities amidst closures and lockdowns due to the pandemic (Enguita et al., 2020). During the pandemic, mothers also took on more home-schooling responsibilities due to school closures (Kasymova et al., 2021). In 2020, mothers spent approximately one hour more per day than their male partners aiding in at-home schooling despite factors related to their employment status (Kasymova et al., 2021). Women are also more likely than their male counterparts to leave paid jobs due to increased domestic responsibilities, despite an increasing trend of women holding higher college degrees (Carnevale et al., 2018).

When considering this information, it is hard to ignore the influence of the American embraced ideology of intensive mothering which states: 1. Children are innocent, sacred gifts and need protection, 2. Child rearing is continuous and when done properly is all-consuming, and 3. Women who are mothers are the primary party to do this kind of work (Hays, 1996). This three-pillar ideology has been widely accepted and became more solidified in societal thought even as movements for gender equity have gained ground due to child rearing practices and restricting options for women that can result in retreating from the workforce and returning to the home (Hartman, 2004). Intensive mothering is in direct conflict with women as mothers seeking full-time employment. One cannot equally serve both full time commitment to work and full-time devotion to the family (Blair-Loy, 2003).

One reason that has been proposed for differences in earned wages and power in the workplace when comparing women and men is that women are more risk averse than men (Charness & Gneezy, 2012). This can have a negative impact on salary negotiations and financial decision making for women (Marks & Harold, 2011). For example, a woman and a man are offered the same job with the same starting salary, but if the man is more likely to negotiate his salary, he may receive higher pay that is reflected in a gender pay difference as an aggregate outcome. In studies of personality differences between men and women, women score higher in trait agreeableness as compared to men, although the perception of women as agreeable may exceed the actual data (Biron et al., 2016; Judge et al., 2012). Maitra et al. (2021) found that although most of the causes of the gender pay gap is are unexplained, women being more risk averse coupled with a lower likelihood of negotiation accounted for 15.5% of the gender wage gap in a Vietnamese sample. These results are compatible with earlier outcomes reported by Gerhart and Rynes (1991) who demonstrated that female Masters of Business Administration (MBA) graduates are likely to obtain lower monetary returns from negotiation than their male counterparts which can result in considerable accumulated differences over time. Motherhood may also further exacerbate these differences. Working mothers are more likely than working fathers to say family obligations interrupted their own career advancement, and may show greater risk aversion than fathers, resulting from the false choice for women between being an “ideal worker” or a “good mother” (Abbott-Chapman et al., 2008; Livingston, 2020).

Any approach that places risk-aversion and agreeableness within the “personality” of women (expressed as an average) is immensely problematic. Ending the analysis here risks blaming women who are the ones disadvantaged by oppressive systems – not the cause of them. This approach is also inconsistent with functional contextualism that is at the heart of contextual behavioral science (Biglan & Hayes, 1996) that emphasizes behavior as it is situated in an context and adopts the same pragmatic assumptions that laid the foundation for women’s suffrage and feminist movements. According to Nelson (2014) in a review of economic literature, there is considerable overlap in studies of risk between men and women exceeding 80%, and studies that evaluated contextual influences suggest that external factors are more predictive of risk-aversion than differences based on one’s gender. Women entering the workforce also expect a mean entry self-pay of about $34,000, whereas men expect $43,000, nearly a $9,000 yearly gap (Heckert et al., 2002). Women also estimate an average of 40% for their asked time off for child-rearing, whereas men expect 4.1% of their time off will be needed for child-rearing (Heckert et al., 2002). In the case of the shared experiences of women, there is ample evidence that social contexts may generate oppressive contingencies that limit access to valued reinforcers for woman, and for mothers, in a patriarchal society (O’Reilly, 2019).

Risk-taking and risk-aversion can be interpreted from within a behavioral economic framework in the form of probability discounting. Probability discounting describes a decrease in the subjective value of a commodity as the odds-against receiving the commodity decreases (Green et al., 2014). Probability discounting becomes suboptimal when the reward exceeds the risk – for example, given the choice between a 60% chance of $1000 (option A) or a 100% chance of $500 (option B), the second choice may be considered suboptimal because repeated choices will result in less reward over time (i.e., 10 choices will on average result in $6000 for option A and $5000 for option B). In monetary probability discounting tasks, concurrent choices are provided between a smaller-certain amount of money versus a larger-probabilistic amount of money, and probability discounting is modelled when participants choose the smaller-certain amount of money at a higher rate as the odds-against receiving the money increases. This relationship has been modeled using Rachlin and colleagues’ (1991) hyperbolic model, where the discounting rate is represented by h in the equation: \(V=A / (1+\) hθ), where V is the subjective value of the reward, A is the undiscounted value of the reward, and θ is the probability of the reward. Risk aversion represents a tendency to overweight the odds-against function resulting in greater values of h (i.e., greater probability discounting; Shead & Hodgins, 2009). In this case, steeper probability discounting is evidence of greater risk aversion, or sensitivity to risk (i.e., likely to choose less money to avoid risk of receiving no money).

A number of contextual factors have been shown to influence discounting rates (Odum et al., 2020) and a series of studies have introduced hypothetical contextual variables to isolate the relative influence of these contextual factors. Dixon et al. (2016) conducted an experimental analysis with nine participants with gambling disorders who completed a monetary delay discounting task under the hypothetical conditions of earning twice or half as much as they currently made in their work. Results showed the greatest discounting rates in the deprived economic condition suggesting that income may operate as a contrived motivating operation. In a similar arrangement, Belisle et al. (2019) recruited 36 college student participants who completed a modified social discounting task (i.e., discounted value of reinforcement for another person as a function of social closeness). Three hypothetical scenarios involved either the participant was facing economic hardship (self), the hypothetical other was experiencing economic hardship (other), or both the participant and the other were facing economic hardship (both). Results showed the greatest discounting rates in the other condition. More recently, Belisle et al. (2022) evaluated probability discounting and isolating or distancing during the coronavirus COVID-19 pandemic, where participants chose to distance or isolate for a number of days as a function of the perceived risk of the pandemic. In the scenarios, participants made choices that were constrained to their own behavior (individual contingency) or assumed their choices would represent the choices of their community (group contingency), and results showed greater discounting in the individual contingency scenarios.

This strategy may help to resolve the correlational nature of research on gender, motherhood, and risk aversion that leads to personality or intrinsic explanations for choices of women and mothers in society. It is not possible to experimentally add or remove one’s parenting status from their context, but it is possible to introduce parenting or non-parenting as a hypothetical contextual variable to determine the relative influence of this whole context on decision making. There are a number of issues when relying exclusively on correlated outcomes that leave open interpretation as to the apparent causal status of motherhood status and risk aversion (Belisle et al., 2021). First is the directionality problem – it may be the case that women who are more risk-averse are also more likely to have children. Second, there may be underlying trait characteristics (i.e., third variables), like agreeableness or compassion for others, that contribute to both risk-aversion and the choice to parent (i.e., both are outcomes of the same cause). Other confounding variables might occur if lower-income families tend to have more children and the link between income and discounting is well documented (Bickel et al., 2015), or that mothers are on average older than non-mothers creating an additional confound. Hypothetically adding or subtracting children from one’s present context in a controlled experimental arrangement confers the advantage of holding all of these other variables constant as the person completing the task in a parent condition is the same person (i.e., same traits, economic status, experiences, etc.) completing the task in the non-parenting condition, allowing for the isolation of motherhood as one context variable that may contribute to risk-aversion and, consequentially, economic disadvantages for women within a broader social context of motherhood.

Therefore, the present set of two experiments provided an initial empirical attempt to isolate motherhood as a broad construct variable on probability discounting. The first experiment was an additive design and involved the recruitment of college students who identified as female and were not currently parents (i.e., non-mothers) who completed a modified probability discounting task in a current control condition and under the hypothetical scenario of having one child. The second experiment was a subtractive design and involved the recruitment of a representative sample of participants who identified as female and were currently parents (i.e., mothers) who completed the same probability discounting task in a control condition and under the hypothetical scenario that they had elected not to have children. The study was designed to allow for future analyses isolating specific events within intersectional experiences of motherhood and covariate analyses with larger and more representative samples.

Experiment 1: Additive Design: Non-Mother Convenience Sample

Methods

Participants

Participants for this initial investigation were recruited from undergraduate psychology courses at a university in the Midwestern United States and represent a convenience sample. Sampling also included participants who were less likely to have children (i.e., college women) than the general same-age population (Finer & Zolna, 2011). This sample also represents individuals preparing, through higher education, to enter the workforce and likely to be impacted in the near future by the contextual contingencies surrounding womanhood and motherhood in workplaces. Therefore, the sample may have been less likely to have experienced directly the contingencies explored in this initial investigation which was a strength of this sampling method. A total of 56 participants were recruited for the study and 43 participants were retained based on the initial criteria in the data analysis phase. All participants identified as females and non-mothers. The ages of the participants ranged from 19 years to 36 years (M = 21.8 years, SD = 3.6 years). Out of the 43 participants, 38 identified as White, three as Black, one as Hispanic, and one as Native American. Participants received extra credit in their course for participating in the study. All research methods were approved by the university’s Internal Review Board prior to the implementation of the study.

Materials and Setting

This study was conducted remotely over zoom during the class period. This was done to ensure the transportability of the research methods to allow for future recruitment of more diverse and representative samples if experimental control was apparent given this arrangement. This study was delivered using Qualtrics (Qualtrics, Provo, UT) that allowed for the delivery of the experiment as well as obtaining data that were stored in a secure university server. No identifying data were collected. Materials included a brief demographics survey and two probability monetary discounting tasks. The first task was completed in a control condition where participants completed the task assuming their life was the same as it was presently (non-mother condition; Child (−)). The second task was completed in an additive condition where participants completed an imaginary scenario where their life was the same except that they had chosen to have a child (mother condition; Child (+)).

The probability discounting task was modified from Estle and colleagues’ (2007) monetary discounting task based on pilot research presented by [blinded for review, 2021]. In the original analyses, differences between Child (−) and Child (+) conditions were apparent within a range of values called the range of ambiguity (near the middle of the odds-against values) and not at either extreme end of the probability discounting function (i.e., extremely low or extremely high odds against). To allow for within subject replication across odds-against values, more concurrent choices were presented within this range of ambiguity. Therefore, participants were provided concurrent choices between an amount of money that was 100% certain (range, $1000 to $100, presented in decreasing values of $100) or $1000 that was probabilistic. The probability values for the larger sum of money were 90%, 75%, 70%, 65%, 60%, and 10%, where the values 75% to 60% were within the ambiguous range based on pilot research. The concurrent choices were identical in both conditions and took the form of the example:

$$100\%\; chance\; of\; \$600$$

or

$$75\% \;chance\; of\; \$1000$$

Figure 1 shows an example of what participants viewed in the probability discounting task. The sequence of the discounting tasks was randomized across participants to counterbalance potential sequence effects. Details specific to both tasks are described below:

Fig. 1
figure 1

Example screen shot of the discounting task as viewed by the participants. Left is viewed on a laptop or desktop and right is viewed through a smartphone application

Full size image

Child (−) Probability Discounting Task (Control Condition)

In the discounting task participants were provided the following instructions to answer the questions so that all participants were responding from their current context:

Think about yourself in your current situation. How old you are, what is your level of education is, your job(s), your current living situation, and your financial stability.

As you answer the following questions, keep all these details about your current self in mind.

Participants then proceeded to answer the concurrent choices as presented in the discounting task.

Child (+) Probability Discounting Task (Experimental Condition)

The experimental condition was designed to present the context of having a child and bringing experiencing the child to the psychological present when completing the discounting task. The first part was developed to hold the age and interests of the child constant across all participants as a variable that could be manipulated in future research. Participants were provided the following instructions:

Think about yourself in your current situation. How old you are, what is your level of education is, your job(s), your current living situation, and your financial stability, and imagine you have a young child. You have a 6-year-old boy named Jimmy and he just started 1st grade. He loves playing with dinosaurs and toy cars and always wants to play outside. His favorite foods include mac and cheese and chicken nuggets, but he hates fruit snacks. He is always full of energy and jokes and constantly makes you laugh. Jimmy enjoys quality time with his friends playing soccer and video games. Jimmy also loves playing games and watching movies with you.

The second component was designed to elicit feelings of the child being present. Participants were given 2-minutes to write in a short answer box what a day at the park with Jimmy would look like. After writing a description of what their day would look like, the participants immediately completed the monetary discounting task.

Dependent Variable and Data Analysis

The primary dependent variable was the indifference point at each odds-against value within the probability discounting tasks. The indifference point was the median value between which participants selected the smaller-certain value and the larger-probabilistic value. For example, if participants selected 100% of $600 but not 100% of $500 when presented concurrently with a 75% chance of $1000, then the median indifference point for that participant at this probability value was $550. Therefore, a total of six indifference points were obtained for each participant. Adhering to previous research on probability discounting, the probability of monetary reward (x-axis variable) was converted to odds against receiving the monetary reward using the formula: \(X=(100-p)/p\), where p represents the probability of receiving the monetary reward. For example, a 75% chance of receiving the money is converted to (100 – 75) / 75, which is equal to 0.33 odds against. Participant data were not retained if they did not answer all of the questions in the discounting task or if they switched on multiple occasions in a probability value as their indifference point could not be concluded.

We used two primary statistical analyses to evaluate results in the current study. Both were completed using Statistica software (VinceStatSoftware, 2008). First, we evaluated hyperbolic functions for probability discounting curves within each condition using the probability discounting function proposed by Rachlin et al. (1991) producing the value h as a free parameter describing the discount rate. In accordance with previous probability discounting measures, we anticipated a goodness of fit index of R2 ≥ 0.90 suggesting that the hyperbolic model retained a good fit under the hypothetical contextual scenarios presented in our current study. Goodness of fit was indexed using the least squares loss function and an initial value of A = 500 based on visual analysis of the data. We then compared discounting rates between the two conditions by calculating Area Under the Curve (AUC) values for each participant using the ordinal transformation proposed by Borges et al. (2016). In this method, the original scaling for the probability discounting measure can be replaced with successive integers. Replacing each probability with an integer results in equal spacing probabilities. This transformation was necessary given the modified probability values used in the present study.

Results

Results from Experiment 1 are displayed in Fig. 2, which shows the median indifference point values and the best fit hyperbolic curve function (top) and the AUC data (bottom). For the Child (−) condition (control), the hyperbolic function produced a best fit h value of 2.81 and a goodness of fit of R2 = 0.95 that is consistent with previous research. For the Child (+) condition, the hyperbolic function produced a best fit h value of 1.37 and a goodness of fit of R2 = 0.98. Therefore, results replicate the hyperbolic discounting function in the Child (+) condition and support a steeper discounting function that is also supported by visual analysis of the median indifference point data in the figure.

Fig. 2
figure 2

Mean indifference points with the hyperbolic function fit to the data (top) and individual AUC data (bottom) with mean, standard error, and raw data plotted. These data represent both conditions.

Full size image

In the AUC comparison, each data point is an individual participant. Visual analysis of the data support intersubject variability that is likely due to variable contexts for each participant in the study and this is consistent across participants. Consistent with the curve fit analysis below, we observed lower AUC values in the Child (+) condition (M = 0.28, SD = 0.13) as compared to the Child (−) condition (M = 0.32, SD = 0.12). AUC is an atheoretical measure that also allows for the use of conventional inductive statistics for comparison across groups or conditions (Myerson et al., 2001). Results of a paired-samples t-test show that these differences were statistically significant (t(37) = −2.30, p < 0.05). We also evaluated the effect size using Cohen’s d for paired sample data. Results supported a small to medium effect size (Cohen, 1988, 1992) of d = 0.35. Given the exploratory nature of the current study, we conducted a power analysis to help guide future research in obtaining more representative samples using G*Power statistical software. The results of the power analysis using the obtained effect size suggest a sample of at least 106 participants may be necessary in future investigations.

Experiment 2: Subtractive Design: Mother Sample

Participants

Participants for the second experiment were recruited online using Amazon’s Mechanical Turk and were paid $8.00 to complete the study. Participants were recruited online to ensure a more representative sample and to constrain the sampling parameters to participants who identified as mothers at the onset of the study. The sample was further restricted to United States residents to remain consistent with the IRB approval. A total of 58 participants completed the probability discounting tasks and 51 were retained based on the inclusion criteria that were the same as in Experiment 1. Participants were recruited from 12 different states. Ages ranged from 26 years to 68 years (M = 41.7 years, SD = 8.8 years) and identified age ranges for children were from 3 years to 34 years. Out of the participants, 41 identified as White, four identified as Black, two identified as Multiracial or Biracial, two identified as Hispanic, one identified as Native American or Alaskan Native, and one identified as Asian or Pacific Islander. All participants identified as female and mothers. Participants were paid to complete the experiment and methods were approved by the Institutional Review Board at the university.

Materials and Setting

Materials were again developed on Qualtrics, and data were stored on a secure university system. Participants completed the study remotely through the M-Turk system. The probability discounting tasks were identical to Experiment 1. The Child (+) was now the control condition as participants were mothers and the Child (−) condition served as the control condition. Small modifications were made to the Child (+) condition to allow for stronger experimental similarity between conditions. The conditions were again randomized to counterbalance potential sequence effects. Details specific to both tasks are described below:

Child (+) Probability Discounting Task (Control Condition)

Participants were presented with instructions and an imagination task in the control condition. The imagination task was added to address the potential confound of time and effort that differed between conditions in Experiment 1. The instructions were designed to orient participants to their current circumstance while completing the discounting task:

Think about yourself in your current situation. How old you are, what is your level of education is, your job(s), your current living situation, and your financial stability.

Participants were then asked to imagine their life with their children with the following instructions followed by a 2-minute window to reflect on their experiences:

What are your responsibilities with caregiving? Think about the time you have for yourself, as well as time dedicated to family. What feelings do you experience daily as a parent? What thoughts come to mind? Please take two minutes to write a brief reflection about these experiences.

As you answer the questions that follow, keep all these details in mind.

After the 2-minutes elapsed, participants were then presented with the discounting task.

Child (-) Probability Discounting Task (Experimental Condition)

The Child (−) condition was developed similarly to the Child (+) condition except the imagination scenario was designed to evoke thoughts about life had the choice been made not to have children. The participants were provided with the same initial instruction followed by the instructions below and 2-minute reflection window:

Imagine a world where you decided not to have any children, and you were not a mother. Think about yourself in this situation. Imagine a day in your life without children. Think about the time you have for yourself, as well as dedicated to family. What feelings might you experience if you had chosen not to be a parent? What thoughts come to mind? Please take two minutes to write a brief reflection about these experiences.

After the 2-minutes elapsed, participants were then presented with the same probability discounting task.

Dependent Variable and Data Analysis

The analytic strategy for Experiment 2 was identical to Experiment 1.

Results

Results from Experiment 2 are displayed in Fig. 3, which shows the median indifference point values and the best fit hyperbolic curve function (top) and the AUC data (bottom). For the Child (+) condition, the hyperbolic function produced a best fit h value of 1.39 and resulted in an R2 = 0.98. For the Child (−) condition, the hyperbolic function produced a best fit h values of 4.00 and resulted in an R2 = 0.97. These results mirror the results from Experiment 1, where both conditions produced curve fit estimates that are consistent with probability discounting without the experimental manipulations. Moreover, the subtractive design suggested that the subtraction of the motherhood context reduced the probability discounting rate. Whereas the Child (−) condition operated as the control condition in Experiment 1, the Child (+) condition operated as the control condition in Experiment 2. In both cases, the Child (+) condition resulted in steeper discounting based on the curve fit analyses for the Child (+) condition.

Fig. 3
figure 3

Mean indifference points with the hyperbolic function fit to the data (top) and individual AUC data (bottom) with mean, standard error, and raw data plotted. These data represent both conditions

Full size image

Consistent with the curve fit analysis, lower AUC values were also evident in the Child (+) condition (M = 0.35, SD = 0.14) as compared to the Child (−) condition (M = 0.49, SD = 0.21). Like in Experiment 1, results of a paired-samples t-test also show that these differences were statistically significant (t(50) = −5.29, p < 0.01). The results of Cohen’s d effect size analysis supported a medium to large effect size of d = 0.76. Using this value, we again conducted a power analysis to guide future research. Because of the larger effect size, results suggested that a sample of at least 25 participants may be necessary in future investigations.

General Discussion

In the United States, 63.1 million people are parents, and the roles that embody the experience of mothers may create a collective context that occasions choice patterns that systematically disadvantage women within a patriarchal society and in male-dominated industries (Budig, & England, 2001; Charness, & Gneezy, 2012), including in the field of behavior analysis (Li et al., 2018; 2019; Martin et al., 2022). Risk-averse response patterns, especially in the context of employment and financial decision making, have been indicated as one of multiple causes for disparities in the experiences and earnings of women broadly and mothers in particular (Kochhar, 2023; Lips & Lawson, 2009), although experimental evaluations of the context of motherhood as a contextual variable has not been explored. The current study offers an initial exploration of a potential method to isolate motherhood as a larger context in order to rule out confounding variables and directionality problems that accompany correlational methods that rely on interpretation and theory. In the current study, a first group of non-mothers was asked to complete a probability discounting task in the hypothetical scenario that they had a 6-year-old male child (i.e., additive design) and discounting rates were compared to their current circumstance as a non-mother control. Results showed greater probability discounting rates in the mother condition. A second group of mothers was asked to complete the same discounting task in the hypothetical scenario that they had elected not to have children (i.e., subtractive design) and discounting rates were compared to their current circumstance as a mother control. Results again showed greater probability discounting rates in the mother condition. These results support this strategy as a potentially viable analytic method and suggest that motherhood is a context that may directly influence risk-aversion and decision making that may appear suboptimal, but is couched in a patriarchal context within which the current study took place.

We intentionally did not include a male sample in the present study because our purpose in this study was not to compare the experiences of mothers to the experience of fathers; rather, the purpose was to isolate the potential influence of the context of motherhood on women in its own right. Importantly, the context of motherhood is not a singular context, and the current results may be used to develop new research questions that seek to explore the various components and intersectional experiences that comprise motherhood as a collective variable. For example, researchers have speculated that contemporary versus conservative beliefs about motherhood and social roles could influence decision-making, such expectations that mothers’ involvement should be “intensive” by prioritizing the needs of the children above other values like career advancement (Blair-Loy, 2009; Hays, 1996; Kobrynowicz & Biernat, 1997). Women who are mothers may earn less than non-mothers when having children causes a loss in job experience, trade of higher risk wages for more “mother-friendly” jobs, and discrimination by employers for being a mother or primary caregiver (Budig & England, 2001). These are experiences that appear to be unique to women as an aggregate group and warrant greater empirical investigation (Lips & Lawson, 2009). Furthermore, although these results do not speak to the specific contextual influences of womanhood or motherhood, they do provide a method through which to investigate these factors. For example, Dannals et al. (2021) demonstrated that backlash against women for engaging in stereotype-incongruent behavior accounted for greater variance in negotiated outcomes for women when strong alternative sources of income were not present. In the context of motherhood, societal expectations of intensive parenting may also lessen alternatives to negotiated agreement, which should be explored in future studies to further dissect the functional context.

In the United States, these additional risks that operate outside of workplace (but affect experiences in the workplace) are exacerbated when laws are not in place to protect the reproductive rights of women (Dobbs v. Jackson) and may even further disadvantage women within contemporary society. The impacts of abolishing reproductive rights of women are multiplied when the context of motherhood has a negative impact on available choices for women given the social expectations around parenting, the burden of which is on mothers in the United States (Lips & Lawson, 2009). The current analytic strategy could, therefore, be used to evaluate the potential impact of legal decisions around parenting and reproduction to better inform the society through contextual behavioral science.

There are several implications that can be taken from the current study to inform future research. Differences in responding across the two scenarios appears to be most pronounced in the range of ambiguity initially shown in pilot research by (Venegoni et al., 2020). Differences in responding between mothers and non-mothers may not be evidence at extreme risks or extreme non-risks but are evident in a zone where decisions are more probabilistic. Salary negotiation provides a real-world example where a successful outcome may result in higher annual earnings, yet failure can result in a loss of job prospect. When success or failure is certain, results suggest that differences between conditions may not be anticipated; however, differences may present when successful negotiation is more uncertain (e.g., between 75% and 60%), that may result in small differences in annual salary that appear within aggregate comparative statistics between men and woman and between mothers and non-mothers. That is, the difference may not be between accepting a $30,000 per year job or a $60,000 per year job; rather, the difference may be seen when accepting a $58,000 contract that is guaranteed versus risking the contract for a salary increase to $60,000. Risks may be greater within families for mothers who are seen as the primary caregiver, where losing the job through negotiation has very real impacts on the wellbeing of children.

Although we observed greater probability discounting rates in the mother condition for both experiments, it is important to note that probability discounting rates were overall greater in Experiment 1 (non-mother sample) than in Experiment 2 (mother sample). Experiences are intersectional and the two experiments should not be directly compared. The first experiment recruited a sample of college women who are generally lower income earners as a population compared to women who are not in college (Livingston, 2020). The relationship between income and probability discounting rates have also been established in correlational research (Green & Myerson, 2004) and experimentally by Dixon et al. (2016). If a quasi-experimental comparison was desired, it would be necessary to recruit a sample of college students who are mothers and college students who are not mothers, although income and status as a non-traditional student may operate as a confound that would need to be controlled for statistically. The online delivery method of both experiments may occasion these kinds of statistical analyses by recruiting larger and more diverse samples to explore potential covariates in the data that were not possible in this exploratory analysis.

Several limitations in the current study suggest that further replication and refinement is needed in future research. The first study was completed with convenience samples of college women at a single Midwestern American university. The sample was also predominantly white. Although this provided a relatively homogenous sample for comparison using a repeated measures design, implications are limited with respect to intersectional experiences of women and mothers. A justifiable criticism of the feminist movement is the centering of voices of White women and mothers that limit the voices of women of color and sexual minorities in defining the shared social stories of womanhood and motherhood. Struggles and oppression of White women do not singularly define the struggles and oppression experienced by all woman, and our sampling method is limited in this same way. It is critically important that this method be considered a preliminary effort to guide future research that includes a more diverse sampling method with a greater number of participants to explore contextual covariates. This same limitation exists in the second experiment even though we employed an online sampling method and paid participants. The platform may not be equally accessible to all potential participants, and we did not make an effort to constrain demographic variables to ensure greater diversity in our sample. Future research should make this effort now that initial findings with a non-representative sample have been achieved.

Another limitation is that outcomes were not compared with potential covariates such as current income, familial wealth, experiences with chronic illness in families, or other intersectional experiences. These variables may mediate the relationship between the context of motherhood and probability discounting as risks are not held equally across all social groups. The sample in the present study was not powered enough for these deeper analyses and exclusively allows for the evaluation of motherhood as the manipulated variable. Future research should recruit a larger sample to allow for exploration of these covariates. Additionally, we did not include the imagination scenario in the control condition in Experiment 1; rather participants were instructed to complete the study as their life was presently (i.e., without children). This created a confound in that the content of the imagination scenario and the event of engaging in an imaginative scenario (regardless of the content) were both concurrent manipulations. We addressed this as a limitation in Experiment 2 and the same directional differences were observed providing confidence that it was the content of the imagination scenario and not its presence or absence that produced significant differences in Experiment 1. Future research should, however, employ a similar control condition to that used in Experiment 2 of the present study.

A third limitation could easily be corrected in future versions replicating this procedure. Specifically, we did not provide an attention check to confirm that participants read the scenarios prior-to completing the discounting tasks; therefore, we cannot be certain that this manipulation was observed in all cases. That said, results are differentiated in both experiments with the same general outcome – steeper probability discounting was apparent in the Child (+) condition relative to the Child (−) condition. Therefore, we might expect that eliminating participants from the data analysis who were not attending to the scenarios would have made the results more saliant if attending did not occur in all cases. Relatedly, participants from the second experiment were recruited from Amazon’s M-Turk, and while data acquired using this method may be less reliable than community samples, the presence of an item asking participants to affirm their attention and honesty may be associated with more reliable responses (Rouse, 2015).

Beyond addressing these limitations, several additional avenues for future research could extend from the present studies. We treated motherhood as a singular collective context variable, whereas there are various components of motherhood that could be further analyzed. For example, parental pressures associated with being a single mother may be different than being a mother in a committed relationship, where greater probability discounting may be evident for single mothers or single income earners. Socioeconomic status may also interact with the context of motherhood, where methods employed by Dixon et al. (2016) could be combined with methods used in the present studies to evaluate main effects and interaction effects within a repeated measures factorial design. Similarly, the parental involvement of the other parent ranging from primary parent to equal parent to uninvolved parent could also influence probability discounting rates to isolate social expectations of women in the role of mothers as a salient feature within the context of motherhood.

Another area for future research would be to explore contextual interventions that could be embedded in workplaces that could protect against social pressures on women and mothers (i.e., developing affirming and supportive workplaces; Chrobot-Mason & Aramovich, 2013; Poduval & Poduval, 2009). One example may be including paid maternity and paternity leave within employment contracts that mitigate some of the financial risks associated with parenting. Currently in the United States, where the present studies took place, paid maternity leave is not guaranteed for female workers and paternity leave as an equal co-parent is not common. These policies can shift gendered dynamics around parenting that may influence the social context of motherhood and dismantle some of the many oppressive systems. Other employment-based solutions may include subsidies for childcare for working parents (Zanoni & Weinberger, 2015), flexible working hours for parents (Chung & Van der Lippe, 2020; Genadek & Hill, 2017), among other solutions (Huang et al., 2022). Because of the emphasis placed on negotiation as one example of the negative impact of risk aversion, future research should also include a negotiation task to align probability discounting outcomes more closely with real-world choices of interest.

Interventions could be centered around mindfulness and values such as Acceptance and Commitment Training for women and mothers. Dixon et al. (2019) found that brief mindfulness interventions can influence momentary delay discounting rates. In this case, mothers may be responding to verbal rules developed through a relational history organized around womanhood and motherhood (e.g., “to be a good mother I must always put my children’s needs first”), where mindfulness may allow for greater awareness of the external contingencies when making critical financial decisions or other decisions that are influenced by this context. Values clarification activities may also help orient participants toward desired sources of reinforcement beyond gendered roles and parenting that could also reduce risk aversion and encourage committed action to approach other sources of valued reinforcement, an example being that family is an important value, but not the only value (Engle & Follette, 2018; Paliliunas, 2021).

Finally, results could be expanded by recruiting fathers and non-fathers as well as parents of non-binary identification. Although the purpose of these studies was to explore womanhood and motherhood specifically, challenges and pressures are also apparent within manhood and experienced by fathers. For example, as of September 2022, 93.1% of incarcerated people identify as men and 78.7% of homicide victims are also men, as reported by the United Nations. In addition, the American Foundation for Suicide Prevention reported that 49,979 people in the United States died by suicide in the year 2020 and men died by suicide 3.88 times more than women who died by suicide in 2020. White males also made up 69.68% of the reported suicide deaths in 2020. Pressures associated with being a primary earner in a family when that family includes children may exacerbate these issues, necessitating experimental analyses that are developed specifically to evaluate these challenges. Patriarchal systems oppress women in society but can also have negative implications for men in a society, and this requires further empirical investigation. For people who identify beyond the gender binary, securing and maintaining a job can be laborious. For instance, there are no current federal laws providing direct legal protections for transgender workers based on gender identity or expression. Those who identify beyond the gender binary experience unemployment at twice the rate compared to the population as a whole (Rathjen, 2013).

In summary, women experience financial disadvantages compared to men in the United States and mothers even more so. A fundamental assumption within both feminist theory and matricentric feminism is that gender and parent roles are socially constructed and evolved throughout a patriarchal history rooted in the systematic oppression of women (Soman, 2009). This is a pragmatic way to approach challenges of oppression by placing influence within the context of womanhood and motherhood instead of some inherent or innate state of being associated with being a woman. This is also consistent with the pragmatic tradition of the contextual behavioral sciences and functional contextualism more broadly. The present studies provided an initial empirical evaluation exploring motherhood as a context variable that can be manipulated, potentially occasioning future research in this area. Supporting women and mothers is critical within the field of behavior analysis and in society more broadly. As noted by Sophia Loren, “When you are a mother, you are never really alone in your thoughts. A mother always has to think twice, once for herself and once for her child.”