Early experiences with math in childcare and school settings or at home with family members lay the foundation for children’s developing math abilities and beliefs (e.g., Eason et al., 2022; Klibanoff et al., 2006; Zhang et al., 2020). Children’s early mathematical abilities, such as counting, recognizing numerals, comparing numbers to say which is more and which is less, and repeating and extending patterns predict their math achievement later in primary and secondary school (Duncan et al., 2007; Rittle-Johnson et al., 2017). Longitudinal studies of math achievement show relative rank-order stability over time, such that children who have lower performance in math in early elementary school tend to show lower performance in math throughout high school (Davis‐Kean et al., 2022; Ribner et al., 2023). Children’s math abilities similarly predict their affective beliefs and attitudes towards math. Young children with higher math skills tend to report liking math more 5 months later (Fisher et al., 2012) and are less likely to experience later math anxiety relative to children with lower math skills (Gunderson et al., 2018; Ramirez et al., 2018). Moreover, after a period of relative stability during elementary school (Wigfield et al., 1997), longitudinal research suggests that older children and adolescents’ interest in math tends to decline over time (Frenzel et al., 2012; Jacobs et al., 2002), while roughly 40 percent of adolescents show patterns of consistently high or increasing math anxiety during secondary school (Ahmed, 2018). There is also some evidence to suggest that these patterns in attitudes are domain-specific rather than domain general and representative across subjects of early learning. For example, some prior research suggests that young children report greater interest in reading compared to math (Nurmi & Aunola, 2005; Wigfield et al., 1997).

These predictive relations suggest that children’s early math learning experiences may play a crucial role in explaining individual differences in their math achievement and anxiety across childhood into adolescence and emerging adulthood. For example, prior theoretical work suggests that early experiences that focus on visual-only presentations of number (rather than combining with additional auditory information) will increase children’s selective attention to and learning of mathematics over time (Bahrick et al., 2020). However, we are unaware of longitudinal research examining the relations between characteristics of math experiences in early childhood and math outcomes in adulthood. This is likely due to the substantial methodological barriers in conducting a longitudinal study spanning over a decade of childhood. In the present study, we instead use a retrospective method to assess the characteristics of adults’ earliest math experiences, taking a first step towards addressing this gap. Our approach is inspired by previous research on adults’ memories of prior math experiences, which has demonstrated associations between math experiences in adolescence and adulthood outcomes (Towers et al., 2017). We build on and extend previous research in adults’ math memories by drawing from research in STEM identity development, which demonstrates consistent, positive associations between characteristics of science experiences during early childhood and having a STEM identity in adulthood (i.e., seeing oneself as a “STEM” or “science person”; Dou et al., 2019).

Little is known about adults’ memories of math experiences in early childhood. Although many adults are unable to remember experiences that occurred before they were 3 to 4 years old (Tustin & Hayne, 2010), people can recall specific past events in detail when they are emotionally salient experiences (McAdams & Bowman, 2001). Similar experiences that prompt similar emotional reactions that are repeated over time may ultimately become self-defining memories, which organize significant experiences around a central emotional theme (John et al., 2022; Singer, 1995). According to the control-value theory of achievement emotions (Pekrun, 2006), students’ emotional states during domain-specific academic tasks form feedback loops that predict both their performance on the task itself as well as their perceptions of personal control and value during future tasks within that domain. For example, a student who experiences positive emotions during a math task is more likely to experience positive outcomes on that task, and in turn, feel more positively and experience more positive outcomes on future math tasks (John et al., 2022; Pekrun, 2006).

Prior research has examined children and adults’ open-ended, narrative descriptions of math experiences, demonstrating associations between the emotional valence of participants’ math narratives and their current math attitudes and beliefs (John et al., 2020; Towers et al., 2017). For example, John and colleagues (2020) asked undergraduate students to describe a turning point in their mathematics learning and found that participants with more negative memories reported higher math anxiety and lower math self-expectancy and task value relative to participants whose narratives were more positive. Moreover, participants with more negative memories were more likely to report that they planned to avoid math in the future (John et al., 2020). These findings provide further evidence that prior experiences of math may relate to long-term attitudes and beliefs, although past work has focused on math experiences that occurred during adolescence or early adulthood (Towers et al., 2017). However, by the time someone enters adolescence and emerging adulthood they have already had many prior years of math experiences, and in turn, likely hold a set of personal affective beliefs about math (e.g., math anxiety, math interest, STEM identity).

To our knowledge, no prior research has examined adults’ earliest memories of math, leaving open questions as to the context and content of these experiences, patterns in their emotional valence, and whether and how they are associated with long-term math attitudes. In addition, although previous work has demonstrated associations between memories of math and participants’ perceived competence in math (Towers et al., 2017), no prior research has examined associations with participants’ math achievement. Identifying characteristics of early learning experiences that are associated with long-term math outcomes may provide an opportunity to support mathematical development early in the lifespan before attitudes and abilities solidify.

The present study extends prior research on math memories to include adults’ earliest memories of learning and their association with math attitudes and achievement. Guided by prior research, we focus on participants in emerging adulthood (aged 18 to 29 years old), as they can use higher-level cognitive skills to reflect on and describe their prior experiences and are more likely than adolescents or older adults to be currently making academic and career choices to seek or avoid math (John et al., 2020). The first aim is to describe the characteristics of emerging adults’ earliest memories of math and reading experiences. Specifically, we examined the nature of the early math and reading experiences (e.g., socially interactive, independent); the type of math or reading activity described; the emotional valence of the experience; and the participant’s approximate age at the time of the memory.

The second aim is to compare features of earliest math and reading experiences to determine whether emerging adults’ memories differed as a function of domain of early learning. Some evidence suggests that young children report greater interest in reading compared to math (Nurmi & Aunola, 2005; Wigfield et al., 1997), thus we hypothesized that participants’ memories of early reading experiences would be appraised more positively than memories of early math experiences.

Our third and final aim was to examine whether characteristics of emerging adults’ early math memories would relate to their current math anxiety, task value, and achievement. In keeping with the control-value theory of achievement emotions (Pekrun, 2006), children’s emotions during math experiences predict both the outcome of their individual experiences (i.e., more or less successful) as well as their future emotional states and math-related outcomes. Specifically, we hypothesized that more positive memories of early math (but not reading) would relate to lower math anxiety and higher task value and achievement, extending prior research that found associations between memories of math from adolescence and early adulthood related to math attitudes and beliefs (John et al., 2020; Towers et al., 2017). This aim allowed us to answer whether the emotional valence of early learning memories are broadly associated with adult attitudes and achievement, or whether the associations are unique within the domain of math.

Participants were recruited from the research subject pool of the Department of Human Development from a large, public university in the northwestern United States in exchange for course credit. We recruited 164 undergraduate students. Three participants were excluded for not completing the focal questionnaire measure in sufficient detail (n = 1) and for describing recent math and reading experiences instead of their earliest memories of math and reading (n = 2). This left an analytic sample of 161 students (M_Age = 19.57, SD_Age = 2.08; 76.9% women, 20.6% men, and 2.5% nonbinary or other; 79.3% White, 9.7% Asian, 5.2% Black or African American, 1.9% American Indian or Alaskan Native, 1.9% Native Hawaiian or other Pacific Islander, and 1.9% biracial or multiracial; 18.6% Hispanic or Latinx; 44.1% freshman or first year, 27.3% sophomore or second year, 20.5% junior or third year, and 8.1% senior or final year of undergraduate enrollment). Participants’ highest level of math coursework completed in high school was 1.2% basic math or pre-algebra, 1.9% algebra 1, 1.9% geometry, 28.6% algebra 2, 39.1% trigonometry or precalculus, 19.9% calculus, 5.6% statistics, 1.2% college preparatory math, and 0.6% linear algebra.

Participants completed all measures as part of a Qualtrics survey administered in an on-campus lab setting with a trained research assistant present. First, participants answered a series of demographic questions. Next, they were asked about their early learning experiences related to math and reading, their math task value, and their math anxiety. The order of presentation for the early learning experiences, math task value, and math anxiety measures was randomized to prevent confounding from order effects. Finally, all participants completed a 10-item math assessment.

To test for potential order effects, we conducted three one-way ANOVAs with task order as the independent variable and math value, anxiety, and achievement as the dependent variables, respectively. There were no significant effects of task presentation order on participants’ math outcomes (anxiety, task value, and achievement). Demographic information on participants’ age and gender identities are presented from available data (n = 150 – 160). Analyses that include measures with missing data used pairwise deletion for missing responses, with maximum ns ranging from 129 to 161.

Participants’ earliest memories of mathematics varied. Most participants were between 4 and 10 years old during their earliest math memory (Figure 1). Nearly all participants (n = 159; 99 percent) were coded as being actively engaged in their earliest math experience. Most participants described an experience that involved numbers (88 percent); only 3 participants described an experience that involved shapes (2 percent). A quarter of participants described an experience involving being timed or having a time limit (25 percent) or practicing math with a worksheet or flashcards (25 percent). Fifteen percent of participants described a math quiz or test.

Most participants (73 percent) included emotion words in their descriptions of their earliest math memory. Figure 2 displays a word cloud summarizing participants’ use of emotion words in their open-ended descriptions of their math and reading experiences. When asked to rate how they felt during their earliest math experience, participants ranged from 1 (very negative) to 7 (very positive), with an average of 4.5 (between neutral and somewhat positive; SD = 1.75). Participants who were younger during their earliest math memories did not have significantly more positive experiences, r(127) = -.14, p = .11. Participants identifying as male did not have significantly more positive early math experiences than participants identifying as female (M_male = 4.62, M_female = 4.52, Welch’s t-test t(45.44) = 0.27, p = .79), nor did participants identifying as female have significantly more positive early reading experiences than participants identifying as male (M_male = 5.83, M_female = 5.40, Welch’s t-test t(69.69) = 1.46, p = .15).

Table 1 summarizes the types of social relationships and interactions that were described in participants’ earliest math and reading memories. Participants who described other people in their earliest math memory who were participating in the activity with them rated their memory as more positive than participants who did not describe someone participating in the activity with them (M = 4.85 and 4.09, respectively; Welch’s t-test t(111.25) = 2.46, p < .05). However, participants who described someone else in their memory who was not participating in the activity did not have more positive ratings of their experience (M = 4.41 and 4.62, respectively; Welch’s t-test t(126.57) = 0.67, p = .50).

On average, participants were younger during their earliest reading experience compared to their earliest math experience, t(160) = 10.28, p < .001 (see Figure 1). Participants also rated their earliest reading experiences as significantly more positive than their earliest math experiences, M_read = 5.5, M_math = 4.5, t(115) = 4.46, p < .001.

Although nearly all participants described other people in their earliest math and reading memories, participants were more likely to describe another person actively participating in their earliest reading memory compared to their earliest math memory, t(160) = 5.45, p < .001. In contrast, participants were more likely to describe another person who was present but not participating in the same activity in their earliest math memories compared to their earliest reading memories, t(160) = 3.41, p < .001. Participants were more likely to describe someone else as a point of comparison or reference for their performance in their earliest math memories compared to their earliest reading memories, t(160) = 4.58, p < .001. In addition, participants’ earliest math memories were more likely to reference a teacher or tutor (t(160) = 2.89, p < .01) or a friend or peer (t(160) = 5.65, p < .001) compared to earliest reading memories, whereas participants’ earliest reading memories were more likely to reference a female caregiver (t(160) = 7.27, p < .001) or male caregiver (t(160) = 2.72, p < .01) compared to earliest math memories.

Table 2 presents descriptive statistics on participants’ current mathematics interest, anxiety, and achievement. There were no significant differences in the current mathematics interest, anxiety, and achievement scores for participants identifying as male compared to participants identifying as female (Welch’s t-test ps > .05).

Participants’ self-reported emotional valence of their early math memories significantly predicted their math task value, anxiety, and achievement as adults. The simple linear regression with math task value regressed on the emotional valence of their earliest math experience was statistically significant, R² = 0.083, F(1, 126) = 11.38, p < .001. Participants with more positive early math memories reported higher math value and interest beliefs (B = 0.210, t(126) = 3.37, p < .001). The simple linear regression with math anxiety regressed on emotional valence was also statistically significant (R² = 0.053, F(1, 127) = 7.13, p < .01), such that participants with more positive early math memories had lower math anxiety (B = -0.094, t(127) = -2.67, p < .01). Furthermore, the simple linear regression with math achievement regressed on emotional valence was statistically significant (R² = 0.051, F(1, 127) = 6.90, p < .01), such that participants with more positive early math memories had higher math achievement (B = 0.256, t(127) = 2.63, p < .01). The emotional valence of participants’ earliest math experiences remained a significant predictor of adulthood math task value, math anxiety, and math achievement in multiple linear regression models controlling for participants’ highest level of high school math coursework (see Appendix B; Scalise et al., 2025S).

In contrast, none of the simple linear regressions of math value and interest beliefs, math anxiety, and math achievement regressed on the emotional valence of participants’ earliest reading experiences were statistically significant, indicating that participants with more positive early reading memories did not have higher math value and interest beliefs, less math anxiety, or higher math achievement as adults (ps > .05). Finally, participants who were older in their earliest math memories had higher current math anxiety (R² = 0.072, F(1, 159) = 12.38, p < .001, B = 0.237, t(159) = 3.52, p < .001), whereas participants who were younger during their earliest math memories tended to have higher current math achievement scores (R² = 0.038, F(1, 159) = 6.33, p < .05, B = -0.478, t(159) = -2.52, p < .05).

The present study has several limitations. First, the study design does not allow for causal inference. Retrospective accounts of learning experiences during childhood are predictive of outcomes in adulthood (e.g., John et al., 2022; Dou et al., 2019), but longitudinal or experimental studies with multi-year follow up measures would provide stronger evidence of a causal relationship between early math experiences and adult math outcomes. It is possible, if not likely, that the association between memories of math experiences and current math attitudes and ability levels is bidirectional. An adult who has low math anxiety, high math task value, and strong math achievement may feel more positively towards math in general and be more primed to remember a positive early math experience than a negative early math experience. Indeed, studies of memory show strong evidence for mood congruence, such that people are significantly more likely to recall memories that match the emotional state of their current experience (Kensinger & Ford, 2020). Future research could examine this with the use of state-based emotion reports (e.g., what emotions do you feel currently, as you are remembering this experience?) and state-based measures of anxiety and math self-efficacy, versus the more generalized trait measures in the present study. People are also more likely to retrieve memories that corroborate their current sense of self and identity, to maintain self-continuity between their past and present experiences (Kensinger & Ford, 2020; Wood & Conway, 2006). Theories of math interest similarly suggest repeating bidirectional associations with math achievement, such that more positive math attitudes promote greater achievement and vice versa over time (Ma, 1997). Moreover, future research should investigate the interaction effects of early math attitudes, including anxiety and task value, and their longitudinal effects on achievement and memories of math experiences (e.g., Ramirez, 2017). In addition, future research could further examine the relations between the emotional valence of participants’ earliest math memories and their math task value, anxiety, or mathematics achievement with a within-person analysis asking participants to describe both a positive and a negative early experience.

Second, the generalizability of our results is limited due to the nature of our sample. Our sample of emerging adults is not representative of national adult demographics nor of emerging adults enrolled in postsecondary education. The majority of our sample identified as female, and all participants were recruited from an introductory social sciences course. In particular, our sample may reflect fewer gender differences in early math and reading experiences, as well as higher rates of math anxiety and lower rates of math task value and achievement than would be expected had we recruited a sample of undergraduate students majoring in STEM fields explicitly, although participants did vary in their highest levels of prior math coursework. Future research should assess the associations between memories of early math experiences and current math outcomes in a more representative sample of adults.

Finally, as noted above, some of the observed differences between participants’ earliest memories of math and reading may be driven in part by the open-ended prompt we used. Future research should examine domain differences in memories of early learning experiences when specific phases of learning are prompted, such as early memories for learning to do math or read independently.

In sum, our results provide additional evidence for the associations between early math learning experiences and later math abilities and affective beliefs. Although a singular experience with math in early childhood may be unlikely to relate to math outcomes in adulthood, similar and repeated early experiences with math may build over time to form the foundation math skills and attitudes. Researchers, educators, and caregivers should continue to promote early math experiences that are positive and engaging for young children, such as informal, playful math activities.