Consistency of adolescent self-report of age at first engagement in five sexual behaviors

Abstract

Self-report data are integral to sexuality research, yet their reliability is often questioned. We leveraged a cohort study of adolescents aged 13-17 years and living in the United States (recruited 2022-2023) to evaluate the consistency of self-reported sexual behavior initiation. Participants were surveyed at baseline, 4-, 9- and 15-month follow-ups. Eligible participants completed at least two surveys (n=1,300). At each timepoint, participants reported the age they first consensually engaged, if ever, in: penetration of the vagina or anus with fingers (digital penetration), oral sex, sex with a toy, penile-vaginal sex, and penile-anal sex. We assigned response patterns as consistent or inconsistent over follow-up. To evaluate whether reports more proximal to the age of initiation were more consistent, we evaluated the proportion of inconsistent responses stratified by whether first engagement was prior to versus during study follow-up. Consistency was high: 81-94%, depending on the type of sex. Nineteen percent of participants had inconsistent reports for digital penetration versus 15% for oral sex, 12% for penetration with toys, 9% for penile-vaginal sex and 6% for penile-anal sex. Reports collected more proximal to initiation were more consistent. In a diverse cohort of adolescents, we report high consistency of self-report data. Initiating data collection prior to first sex may improve consistency.

Introduction

Early age of initiation of penile-vaginal sex is considered an important indicator of future sexual health because it is associated with reduced use of various contraceptive methods (hormonal, condoms, etc.) (Finer & Philbin, 2013) and an increased risk of sexually transmitted infection (STI) acquisition (Kaestle et al., 2005; Tu et al., 2009). Despite its importance, measuring sexual behavior, particularly the age of behavior initiation, poses challenges due to its private nature (Catania et al., 1990). Prior research has evaluated testing for STIs (DiClemente et al., 2011), prostate-specific antigens (Gallo et al., 2006), and Y-chromosome genetic material (Ghanem et al., 2007) as potential objective markers of sexual behavior. However, these markers have primarily been used in adult populations, are limited in their ability to identify a full range of sexual behaviors, and would not necessarily identify age of behavior initiation. As such, self-report data continues to play a central role in understanding adolescent sexual behavior due to the lack of feasible or ethical objective assessments.

A large body of research has explored issues of reliability of adolescent self-reported sexual behavior. Studies have involved a range of populations, including adolescents in the United States (U.S.) (Alexander et al., 1993; Glassman et al., 2023; Hearn et al., 2003; Rodgers et al., 1982; Tu et al., 2009; Upchurch et al., 2002), Nigeria (Tenkorang, 2021), Kenya (Beguy et al., 2009), Tanzania (Plummer et al., 2004), and South Africa (Flisher et al., 2004; Palen et al., 2008). There has been considerable heterogeneity in the duration of follow-up to test reliability, from ~2-3 weeks (Flisher et al., 2004; Hearn et al., 2003) to a year (Beguy et al., 2009; Tenkorang, 2021; Upchurch et al., 2002) or two years (Palen et al., 2008). Ever engaging in penile-vaginal sex (yes/no) is the most commonly studied measure (Beguy et al., 2009; Flisher et al., 2004; Glassman et al., 2023; Palen et al., 2008; Plummer et al., 2004; Rodgers et al., 1982; Tenkorang, 2021; Upchurch et al., 2002). Studies using that measure show substantial variation in consistency of reporting, ranging from 100% (Hearn et al., 2003) to 60% (Palen et al., 2008) of participants consistently reporting engagement over time. Among those studies, inconsistencies in reporting tended to increase with longer follow-up duration. Fewer studies have evaluated the consistency of reported age at first sex. Those that have, have mostly focused on age of first penile-vaginal sex (Alexander et al., 1993; Beguy et al., 2009; Hearn et al., 2003; Tu et al., 2009), finding low (33%% (Alexander et al., 1993)) to high (94% (Tu et al., 2009)) consistency. One study reported the reliability of reported age of first oral sex among Hispanic adolescents, reporting high reliability (Jerman et al., 2015).

The predominant focus on penile-vaginal sex, as opposed to the wider breadth of sexual behaviors typical in adolescence, including vaginal and anal penetration with fingers, and oral sex, is a limitation of the current literature (Dalenberg et al., 2018). Consideration of sexual behaviors beyond penile-vaginal sex can generate insights about adolescent sexual development (de Graaf et al., 2009) and improve the applicability of research findings to gender and sexual minority adolescents, who may be disproportionately excluded by research that focuses solely on penile-vaginal sex.

Current Study

To facilitate research efforts into sexual behaviors beyond penile-vaginal sex, our aim was to evaluate the consistency of self-reported age of initiation of five different sexual behaviors: penetration of the vagina or anus by a finger (digital penetration), oral sex, sex involving penetration with a sex toy, penile-vaginal sex, and penile-anal sex. We used data from an online cohort of adolescents aged 13-17 years at baseline. Individuals were asked about age at initiation of sexual behaviors on 4 surveys spanning 15 months of follow-up.

Method

Participants

QuaranTeen Health is an online, longitudinal study designed to assess the potential influence of the COVID-19 pandemic on U.S. adolescent sexual behavior and relationships. Recruitment occurred between April 2022 and June 2023 using advertisements on social media websites. Participants were eligible if they lived in the U.S., could complete a survey in English or Spanish, were aged 13-17 years and had not completed high school. Youth who identified as sexual and/or gender minority were purposefully oversampled. Eligible participants provided informed assent and completed an assessment of their capacity to assent which comprised questions aimed at confirming their understanding of study procedures. The study protocol was reviewed and approved by Pearl IRB. A waiver of guardian permission was granted.

Following assent, participants completed a baseline survey, designed to take approximately 40 minutes to complete (mean = 38 minutes, standard deviation [SD] ±10). Participants were invited to complete follow-up surveys 2-, 4-, 6-, 9-, 12-, and 15-months after baseline. As the study progressed, we increased incentive amounts and added additional bonus incentives to encourage higher response rates. Specifically, participants received $15 to complete the baseline, 2-, 4-, and 6-month surveys and $20 to complete the 9-, 12- and 15-month surveys. We also offered participants an additional $5 in later waves for completing surveys within two days of launch. Participants could choose an Amazon gift card, a donation to a pre-specified list of non-profit organizations, or to forgo the incentive.

We used a standardized protocol to detect fraudulent responses. Participants were asked to provide their age in the screener and their date of birth in the baseline survey, which were compared. Respondents who provided inconsistent data were paused in the survey and were unable to complete it until they spoke with research staff to resolve it. Participants provided their date of birth at each follow-up, which was compared to the baseline report. Those who reported inconsistent data had to talk with research staff before being allowed to continue in the survey. Participants who could not provide an age and date of birth that aligned, or regularly reported inconsistent data across surveys were flagged for removal from the study.

Additionally, study staff cross-referenced participants’ IP addresses on a third-party IP fraud check website (Scamalytics Ltd, 2024). If an IP address had a fraud score ≥60%, the participant was flagged for further review, which included investigation of their social media profiles to confirm they appeared to be an adolescent and/or contact by text messaging or telephone to confirm their date of birth. Participants who could not confirm these details were also flagged for removal. Study staff discussed each participant who was flagged for removal with the Principal Investigator, including detailing the observed inconsistencies. All participants flagged for removal were removed after discussion with the Principal investigator.

We incorporated several study design elements to encourage valid reporting, including beginning the survey with questions asking participants to confirm they were in a location in which they felt safe and comfortable answering questions honestly (Ybarra et al., 2008). To improve question clarity, we used inclusive language and were highly specific in questions about sexual behavior (e.g., ‘having sex where a penis enters a vagina’ as opposed to ‘having sex’).

A total of 2,077 participants completed the baseline survey. We discarded 274 (13%) as being fraudulent, yielding 1,803 participants with presumed valid baseline questionnaire data. We excluded 502 (28%) participants who did not provide data for at least two time points (i.e., did not complete any follow-up surveys), and 1 participant who declined to respond to any questions related to the age of initiation for all behaviors across all surveys. Excluded participants were not appreciably different in their demographic characteristics or baseline sexual activity than included participants (Supplemental Table 1). Supplemental Figure 1 details sample creation from the initial click on the study advertisement to the final analytic sample.

There was participant attrition over follow up. Of the 1,300 participants in the analytic sample, 1,147 (88%) were in the sample at month 4, 1,029 (79%) were in the sample at month 9, and 1,006 (77%) were in the sample at month 15. Participants could skip a wave and return for a subsequent wave.

Measures

We collected demographic characteristics via self-report on the baseline survey. Age was calculated based on participant birth date. Participants reported their sex assigned at birth using the question: “What sex were you assigned at birth on your original birth certificate?” They could select male, female, or intersex. Gender identity was captured using responses to the question: “Which of the following best describes your gender identity?” Participants were provided a list of 11 options from which they could select all that applied (e.g., cisgender girl, cisgender boy, transgender girl, transgender boy, non-binary, genderqueer). Those who chose “Another identity” were given the opportunity to write in the identity they preferred. Sexual orientation was similarly collected using a ‘check all that apply’ list in response to the question: “Which of these commonly used terms would you use to describe your sexual identity or sexual orientation?” Participants also self-identified their racial identity by checking all that applied in response to the question: “What is your race?” Hispanic ethnicity was captured using the question: “Are you of Spanish or Latino origin, such as Latin American, Mexican, Puerto Rican, or Cuban?” Participants provided the ZIP code where they lived, which was used to determine their state of residence.

On baseline, 4-month, 9-month, and 15-month surveys, participants responded to the following series of questions related to five sexual behaviors: “How old were you the first time you had sex with someone else that involved a finger going in the vagina or anus when you wanted to?” (hereafter referred to as “digital penetration”); “How old were you the first time you had oral sex (we mean stimulating the vagina or penis with the mouth or tongue) when you wanted to?” (hereafter referred to as “oral sex”); “How old were you the first time you had sex with someone else that involved a sex toy or other object going in the vagina or anus when you wanted to?” (hereafter referred to as “penetration with toy”); “How old were you the first time you had sex that involved a penis going into a vagina when you wanted to?” (hereafter referred to as “penile-vaginal sex”); and “How old were you the first time you had sex that involved a penis going into an anus when you wanted to?” (hereafter referred to as “penile-anal sex”) (Ybarra et al., 2016). Participants could input a numerical age in years (bounded by a minimum of 5 and a maximum of their current age) or select one of two check boxes: “I’ve never had this type of sex” or “Decline to answer.”

Procedure

We grouped participants into one of eight mutually-exclusive categories for each behavior, four of which we considered ‘consistent’ and four ‘inconsistent. ‘Consistent’ categories included: 1) participants who reported that they had never engaged in the behavior for all four surveys, 2) participants who reported the same age of initiation for the behavior across all four surveys, 3) participants who were missing data from one or two surveys, but all available data were consistent, and 4) participants who reported an age of initiation in a follow-up survey after previously reporting no engagement and were consistent in subsequent follow-ups (i.e., initiators). ‘Inconsistent’ categories included: 1) participants reporting never engaging in a behavior after previously reporting an age of initiation, 2) participants reporting different ages of initiation across follow-up, 3) participants reporting both never engaging in a behavior after previously reporting an age of initiation and different ages of initiation over follow up, and 4) participants reporting an age of initiation younger than their baseline age after reporting no engagement at baseline.

Data were missing due to both survey non-response and “Decline to answer” responses, which were recoded as missing. Participants who were missing data from 3 or 4 survey waves for a given behavior (8.2% for digital penetration, 8.1% for oral sex, 7.9% for penetration with toy, 7.9% for penile-vaginal sex, and 8.3% for penile-anal sex) were excluded from analyses of the behavior in question, but retained for analyses of other behaviors, if applicable.

Data analysis

For each sexual behavior, we calculated the distribution of consistency categories for the full sample, as well as stratified by age. Based on sample distribution, 13- and 14-year-olds were combined into one category. We additionally evaluated consistency only among those who reported an age of initiation in at least one wave, because consistently reporting an age of engagement may be more difficult than recalling never engaging in an activity. We conducted two analyses to assess whether consistency was higher when age of initiation was assessed more proximal to the first occurrence of the event. First, we evaluated the prevalence of inconsistent reports among participants who reported engagement in each behavior prior to baseline versus participants who initiated during follow-up. Second, among those who reported initiation prior to baseline, we stratified consistency categories by the number of years between the reported age of initiation and baseline age, categorized as: the same as baseline age, 1 year prior to baseline age, and >1 year prior to baseline age, based on the sample distribution.

To assess whether inconsistent reports clustered within participants, we summed the number of behaviors for which participants had inconsistent reports.

Following the guidance of the American Statistical Association (Wasserstein & and Lazar, 2016) and the International Committee of Medical Journal Editors (International Committee of Medical Journal Editors, 2025), we interpreted our findings based on the direction and magnitude of observed estimates, and not binary statistical significance testing.

Results

Baseline characteristics for the 1,300 participants in the analytic sample are presented in Table 1. The mean age of our sample was 15.7 years (standard deviation 1.2). About half of the sample identified as a cisgender girl, with a quarter identifying as a cisgender boy and twelve percent as non-binary, pansexual, queer or genderfluid. The most prevalent sexual orientation identity was heterosexual (45%), followed by multiple identities (13%), and questioning/unsure (13%). Seventy percent of the sample identified as White race, 11% as multiracial, 7% as Black and 6% as Asian. A quarter identified as Hispanic. Participants lived in every U.S. state, as well as the District of Columbia and Puerto Rico.

Table 1.

Baseline characteristics of the sample (N=1,300)

Characteristics	N (%)
Baseline age, years
13	58 (4.5)
14	159 (12.2)
15	287 (22.1)
16	379 (29.2)
17	417 (32.1)
Sex assigned at birth
Female	889 (68.4)
Male	397 (30.5)
Intersex	14 (1.1)
Sexual orientation identity
Gay/lesbian	71 (5.5)
Bisexual	155 (12.1)
Pansexual	87 (6.8)
Heterosexual	577 (45.0)
Questioning/unsure	170 (13.3)
Selected multiple identities	170 (13.3)
Other identity	52 (4.1)
Missing	18
Gender identity
Cisgender girl	647 (49.9)
Cisgender boy	318 (24.5)
Transgender girl	18 (1.4)
Transgender boy	32 (2.5)
Non-binary/pansexual/queer/genderfluid	154 (11.9)
Something else*	129 (9.9)
Missing	2
Race
White	843 (69.0)
Black	85 (7.0)
Asian	73 (6.0)
Hawaiian/Other	2 (0.2)
Native American/Alaska Native	15 (1.2)
Multiracial	141 (11.5)
Other	63 (5.2)
Missing	78
Hispanic
Yes	321 (25.0)
No	964 (75.0)
Missing	15
Engagement in sexual behaviors
Fingering	540 (41.7)
At baseline	342 (63.8)
Initiated during follow-up	194 (36.2)
Indeterminate	4
Missing/Declined to answer at all 4 surveys	5
Oral sex	508 (39.2)
At baseline	318 (63.9)
Initiated during follow-up	180 (36.1)
Indeterminate	10
Missing/Declined to answer at all 4 surveys	4
Sex with toys	247 (19.0)
At baseline	130 (53.1)
Initiated during follow-up	115 (46.9)
Indeterminate	2
Missing/Declined to answer at all 4 surveys	2
Vaginal sex	358 (27.6)
At baseline	217 (61.5)
Initiated during follow-up	136 (38.5)
Indeterminate	5
Missing/Declined to answer at all 4 surveys	5
Anal sex	146 (11.3)
At baseline	67 (47.2)
Initiated during follow-up	75 (52.8)
Indeterminate	4
Missing/Declined to answer at all 4 surveys	4

^*Includes agender, questioning/unsure, and ‘another gender identity.’

The percentage of the cohort reporting specific types of sexual behaviors varied: Forty-one percent of participants reported an age of initiation for digital penetration either on the baseline survey or in at least one follow-up survey, compared with 39% for oral sex, 28% for penile-vaginal sex, 19% for penetration with toys, and 11% for penile-anal sex. The average age of initiation (calculated using the age first reported by respondents if it was reported across multiple surveys) was 15.3 years (SD 1.6) for digital penetration, 15.3 years (SD 1.8) for oral sex, 15.4 years (SD 1.6) for penetration with toys, 15.6 (SD 1.4) for penile-vaginal sex, and 15.8 years (SD 1.5) for penile-anal sex.

Response consistency was high but varied across behaviors in the entire sample (Table 2) and only among those reporting behavior engagement (Table 3). Most participants either consistently reported not engaging in each sexual behavior or had consistent reports with missing data at some waves. Nineteen percent of participants had inconsistent reports for digital penetration, compared with 15% for oral sex, 12% for penetration with toys, 9% for penile-vaginal sex, and 6% for penile-anal sex. Inconsistency was higher when analyses were restricted to those reporting behavior initiation: 45.8% for digital penetration, 38.6% for oral sex, 61.9% for penetration with toys, 30.6% for penile-vaginal sex and 47.5% for penile-anal sex. In age-stratified analyses, inconsistency increased with age for all behaviors (Supplementary Table 2).

Table 2.

Consistency of reported age of initiation of 5 different sexual behaviors over 4 surveys (N=1,300)

	Digital penetration (N=1,194)	Oral sex (N=1,195)	Penetration with Toys (N=1,198)	Penile-vaginal sex (N=1,197)	Penile-anal sex (N=1,192)
Consistency categories	N(%)	N(%)	N(%)	N(%)	N(%)
Consistent	965 (80.8)	1012 (84.7)	1055 (88.1)	1095 (91.5)	1126 (94.5)
Consistent reports of no engagement in all 4 surveys	390 (32.7)	409 (34.2)	542 (45.2)	482 (40.3)	588 (49.3)
Consistent age of initiation reported in all 4 surveys	71 (6.0)	79 (6.6)	13 (1.1)	61 (5.1)	10 (0.8)
Reported an age of initiation in a follow-up survey after previously reporting no engagement, and were consistent in subsequent follow-ups (initiators)	75 (6.3)	91 (7.6)	14 (12)	65 (5.4)	32 (2.7)
Consistent reports in all 4 surveys except for sporadic missingness	429(35.9)	433 (36.2)	486 (40.6)	487 (40.7)	496 (41.6)

Note: Participants who were missing data from all 4 or 3 surveys for a specific behavior were excluded from analyses of those behaviors.

Table 3.

Consistency of reported age of initiation of 5 different sexual behaviors over 4 surveys among those reporting an age of engagement at any wave

	Digital penetration (N=500)	Oral sex (N=474)	Penetration with Toys (N=231)	Penile-vaginal sex (N=333)	Penile-anal sex (N=139)
Consistency categories	N(%)	N(%)	N(%)	N(%)	N(%)
Consistent	271 (54.2)	291 (61.4)	88 (38.1)	231 (69.3)	73 (52.5)
Consistent age of initiation reported in all 4 surveys	71 (14.2)	79 (16.7)	13 (5.6)	61 (18.3)	10 (7.2)
Reported an age of initiation in a follow-up survey after previously reporting no engagement, and were consistent in subsequent follow-ups (initiators)	75 (15.0)	91 (19.2)	14 (6.1)	65 (19.5)	32 (23.0)
Consistent reports in all 4 surveys except for sporadic missingness	125 (25.0)	121 (25.5)	61 (26.4)	105 (31.5)	31 (22.3)
Inconsistent	229 (45.8)	183 (38.6)	143 (61.9)	102 (30.6)	66 (47.5)

Note: Participants who were missing data from all 4 or 3 surveys for a specific behavior were excluded from analyses of those behaviors.

Figure 1 displays the percentage of each type of inconsistent response, among those with inconsistent responses. The most common forms of inconsistency varied across behaviors. For digital penetration, oral sex, and penile-vaginal sex, the most frequent type of inconsistent report was reporting different ages of initiation over follow-up. For penetration with toys and penile-anal sex, the more common inconsistency was reporting one had not had that type of sex after previously reporting an age at first sex.

Figure 1.

The prevalence of each type of inconsistent response among those who had inconsistent response patterns for each behavior

Reports that occurred closer in proximity to the age of initiation appeared to be more consistent. When we stratified the sample by whether participants had engaged in sexual behaviors for the first time before or during the study, the prevalence of inconsistent responses was higher among those who engaged prior to baseline for each sexual behavior assessed (Figure 2). When we evaluated consistency by years between reported age of initiation and baseline age, inconsistent responses were more common among participants reporting initiation >1 year prior to baseline, compared to the same age as baseline, except for penile-vaginal sex, which was generally consistent (Supplementary Table 3).

Figure 2.

Prevalence of inconsistent responses stratified by whether participants reported initiation of a behavior prior to baseline versus during the study

Inconsistent reports clustered within participants. Among those with any inconsistent reports (N=347, 27% of the total sample), 40% had inconsistent reports for only a single behavior, while 30% inconsistently reported two behaviors, 17% inconsistently reported three behaviors, 8% inconsistently reported four behaviors, and 5% inconsistently reported all 5 behaviors.

Discussion

In this diverse national cohort of 1,300 adolescents 13-17 years of age, we report a relatively high consistency of self-reported age of first engagement in five sexual behaviors: from 82% to 95%, depending on the behavior. The high consistency was driven in part by a large proportion of participants who consistently reported never having had that type of sex across all surveys. When analyses were limited to those who reported behavior engagement, consistency was lower. Further, for all behaviors, we found that the prevalence of consistent responses was higher among participants who reported having sex for the first time during the study compared to those who reported having had sex for the first time before baseline. For all behaviors except for penile-vaginal sex, the prevalence of consistent reports decreased as the time between reported age of initiation and the survey date increased.

Our findings align with, and build on, prior reports. In a study of South African adolescents, participants who reported having engaged in penile-vaginal sex prior to baseline were more than twice as likely to subsequently report never having engaged in penile-vaginal sex by the 5^th wave of follow-up (2 years later), compared with those who reported first sex during follow-up (Palen et al., 2008). In our study, we observed a similar pattern across all five behaviors, although our overall prevalence of inconsistent reports of penile-vaginal sex was lower than the South African study. We also observed that inconsistency was more common among older participants and participants for whom more time had elapsed between the reported age of initiation and study baseline. One explanation for these findings is that misremembering an exact age becomes more common as time from the event passes. Accuracy may be improved by data capture more proximal to the initial event. In the case of inconsistent reports across follow-up, using the first-reported age may increase the likelihood of obtaining accurate data, as has been done previously (Boisvert et al., 2017). Another possible explanation is the possibility that inconsistency becomes more common as participants complete more assessments. Participants who reported sexual experience at baseline necessarily had to report the same age over a larger number of follow up surveys than those who initiated under observation. This would also explain why younger participants were more consistent, as they were less likely to engage in sexual activity and therefore had fewer opportunities for inconsistent reports.

Our findings are unique in that we evaluated multiple sexual behaviors, as opposed to only penile-vaginal sex. The percentages of consistent reports differed across behaviors, with penile-anal sex having the highest percentage of consistency and digital penetration having the lowest. Among initiators, penile-vaginal sex had the highest percentage of consistency and penetration with toys had the lowest. It is possible that consistency is related to the perceived salience of a behavior (Goldberg et al., 2014), and adolescents may be more likely to remember a behavior such as penile-vaginal sex versus digital penetration or oral sex. Increased salience of penile-vaginal sex may explain our findings that inconsistent reports of penile-vaginal sex did not increase with more time since reported initiation, and that penile-vaginal sex had the highest consistency when we only evaluated those who initiated behaviors (70%). These findings align with a study of Hispanic adolescents in the U.S. that found that reliability of reports of age at first penile-vaginal sex was slightly higher than age at first oral sex (Jerman et al., 2015). Another potential explanation for differences in consistency across behaviors is the possibility that the first time someone engages in each behavior might become less salient the more they engage in that behavior. We did not ask participants to estimate the total number of times they have engaged in each behavior, so we were unable to evaluate this hypothesis. We also found that, among participants with inconsistent responses, the most common type of inconsistent response varied by behavior. For digital penetration, oral sex, and penile-vaginal sex, inconsistent reports were more likely to be people reporting discrepant ages of initiation across survey surveys. However, for sex with a toy and penile-anal sex, inconsistent reports were more likely to be participants reporting not having had that sex after previously reporting that they had it. It is possible that forces influencing the consistency of reports are not equivalent for all sexual behaviors and that participants perceive some behaviors as more stigmatized or risky to disclose than others. As research efforts expand to include sexual behaviors beyond penile-vaginal sex, consistency of reports of all sexual behaviors should be continually evaluated.

Our study expands the existing literature in other important ways. Our questions were designed to address noted limitations of commonly used questions to assess adolescent sexual behavior. Specifically, we asked explicitly about consensual sexual activity, as opposed to leaving ambiguity about whether participants should report forced sexual activity (O’Sullivan, 2008). We also clearly defined and delineated between various sexual behaviors, as opposed to using the ambiguous term ‘sex’ or ‘sexual intercourse.’ Additionally, we administered questions via online survey, not face-to-face interviews, to facilitate reporting of potentially sensitive information (Newman et al., 2002) and included questions asking whether participants felt they were in a safe place and able to answer the survey questions honestly (Ybarra et al., 2008). These question modifications and survey approaches may explain our generally higher rate of consistent responses than prior studies.

Our findings have some practical implications for researchers. For all behaviors except for penetration with toys and penile-anal sex, reporting different ages of initiation was the most common form of inconsistency, as opposed to reporting never engaging in a behavior after providing an age of engagement. This implies that participants may struggle to remember an exact age but could be more likely to accurately report whether they had ever or never engaged in a behavior. If age of engagement is not specifically necessary for a research question, then a binary yes/no question may reduce inconsistency concerns. Similarly, if a research question can be adequately addressed using an age range, as opposed to an exact age of initiation, asking for responses within a range may help reduce participant cognitive burden while determining their answer. Additionally, participants’ responses were less likely to be inconsistent if they initiated behavior under observation, which could be due to proximity to the event and/or the fact that there were fewer follow up surveys in which to provide discrepant ages. Prospective data collection of behavior initiation has the potential to improve data accuracy. The frequency with which participants reported different ages of initiation, even though our sample was comprised of adolescents for whom behavior initiation had occurred relatively recently, raises concerns that retrospective capture at age of behavior initiation from adult participants may be quite vulnerable to misclassification. Additionally, studies with longer follow up, or a more sexually active population, may be at a higher risk of inconsistent responses.

Our study has important limitations. Like most longitudinal studies, we observed participant attrition over time. We included participants in the current analyses if they contributed data for at least one follow-up survey and assigned consistency of reports using available data. However, missing data could be considered a form of inconsistency that could be explored in more depth. Our sample was recruited via social media advertisements that talked about ‘making a difference’ and ‘having your voice heard.’ Thus, although the advertisements were purposefully vague to reduce response bias, our sample may not reflect all youth, particularly those who do not use social media, or those who do but do not click on advertisements. While we captured a range of sexual behaviors, we did not have data on other common behaviors including kissing or manual stimulation of a penis or clitoris. It is also important to clarify that consistency does not necessarily mean accuracy, and that participants could have misunderstood the behaviors referenced in the questions. We could not validate participant reports of engagement in sexual behavior. Finally, although the relatively low levels of sexual activity in our sample reflect recent national trends towards declining sexual activity among youth (Herbenick et al., 2022), we do not know with certainty how our findings might generalize to other populations.

Despite these limitations, our study offers important and novel insights into the validity of adolescent self-reported sexual behaviors. We report high consistency of reports of age of initiation of a variety of sexual behaviors in a diverse, contemporary cohort of adolescents recruited via the internet. Participants who initiated behavior during the study period had more consistent reports than those who reported having had sex for the first time prior to baseline, suggesting a potential role for prospective data collection in improving data accuracy. Overall, our findings align with prior studies that evaluated the consistency of age of initiation of sexual behaviors (Beguy et al., 2009; Hearn et al., 2003; Tu et al., 2009) in reporting generally high consistency over follow-up, and extends these findings to sexual behaviors beyond penile-vaginal sex.

Data Availability Statement

The data that support the findings of this study are available from the senior author, KMN, upon reasonable request.