Validation of the dispositional adult hyperfocus questionnaire (AHQ-D)

Abstract

Hyperfocus (HF), or intense, deep concentration on a task, has gained significant research attention in recent years, particularly in regard to clinical populations such as Attention-Deficit/Hyperactivity Disorder (ADHD). The present work aims to provide validation of the 12-item dispositional adult hyperfocus questionnaire (AHQ-D) as a quantitative metric of HF in adults. We preregistered the study design and hypotheses. We administered the AHQ-D and several additional questionnaires to 347 adults (mean ± SD age: 33 ± 11 years; 47% female). Exploratory factor analysis revealed high factor loadings (0.57–0.81) on a single HF factor; item response theory analysis suggested that the questionnaire items had high discrimination and covered a wide range of responses; and we report strong internal consistency metrics (Cronbach’s alpha 0.93, mean split-half reliability 0.93). Replicating our previous work, HF was positively correlated with Conners’ Adult ADHD Rating Scale (CAARS) scores (r(345) = 0.53), suggesting that HF may be related to ADHD traits (though in this sample we did not specifically recruit individuals with ADHD). The AHQ-D demonstrated the hypothesized convergent validity; HF on the AHQ-D was positively correlated with HF measured using a different HF scale (r(344) = 0.69), as well as flow (r(345) = 0.12) and mind wandering (r(345) = 0.39) scores. AHQ-D HF scores showed a weak negative correlation with grit (r(345) = − 0.29). Though there was a weak negative correlation between HF and social desirability response tendency (r(345) = − 0.24), suggesting that those who care more about what others think may report less HF, there was no relationship between HF and extrasensory perception beliefs (r(345) = 0.01), suggesting that participants were not simply biased in their response tendencies. Taken together, we demonstrate strong scale metrics for the AHQ-D, the expected convergent validity, and a general lack of response bias, in addition to replicating our previous association of HF with ADHD traits. We suggest that the AHQ-D can be confidently used in future work as a valid way to measure HF in adults.

Introduction

Hyperfocus (HF), or the state of intense and prolonged concentration on a specific task or topic, has garnered significant research interest in recent years. HF has been reported, mostly anecdotally, in various clinical populations including attention-deficit hyperactivity disorder (ADHD)^1–3 and autism spectrum disorder (ASD)⁴. Only recently have steps been taken to develop approaches to quantitatively measure HF tendencies. Ozel-Kizil et al. developed one questionnaire to measure HF in neurotypical adults and those with ADHD^5,6, though this questionnaire suffered from several limitations (e.g., items focused primarily on negative aspects of HF; the published English translation included several items with confusing wording for native English speakers). Subsequently, our group published the adult hyperfocus questionnaire (AHQ)⁷, which has since been utilized to quantify HF in both the general population^8,9 and in various clinical populations, including those with ADHD^10–12.

Our original AHQ⁷ included six parts: four 12-item subscales (dispositional HF and HF related to school, hobby, and screen time activities), one 18-item subscale with descriptions of various scenarios in which HF might occur, and a short-answer section. In the present work, we aimed to shorten and validate the original AHQ, with the goal of making the questionnaire more practical for use in research and clinical settings. Based on our systematic scale development process (see the Methods "Scale development approach" section), we decided to move forward with validation of only the 12-item dispositional subscale (AHQ-D) as a short form of the original 6-part questionnaire. Thus, the primary goal of the present work is to validate the AHQ-D as a quantitative metric of HF in adults.

As described in detail in our prior work⁷, the AHQ-D is comprised of 12 questions, with two questions asking about each of the six “dimensions” included in our working definition of HF:

Hyperfocus: a state of heightened, intense focus of any duration; this state may include the following qualities: timelessness, failure to attend to the world, ignoring personal needs, difficulty stopping and switching tasks, feelings of total engrossment in the task, and feeling “stuck” on small details⁷.

This definition of HF and the resulting AHQ-D items were originally developed based on semi-structured interviews with individuals with ADHD, as outlined in our previous work⁷.

Here, we present psychometrics for the AHQ-D including factor structure, Item Response Theory (IRT) analysis, and metrics of internal consistency. We then examine the convergent validity of the AHQ-D; we predicted that higher AHQ-D scores (higher HF) would correlate with higher HF scores on the Ozel-Kizil HF scale⁵, higher dispositional flow scores^13,14, and greater spontaneous mind wandering¹⁵. We selected flow (i.e., full immersion in an intrinsically enjoyable activity)¹⁶ as a secondary convergent metric, as both HF and flow involve allocating intense focus to a task, and because we found a positive correlation between dispositional HF and flow in our prior work⁷. We selected mind wandering as an additional secondary convergent metric because we view both HF and mind wandering as facets of poor attentional control, as supported by our previous finding that a higher susceptibility to mind wandering was associated with a stronger tendency to engage in HF⁸. Next, we assess divergent validity, predicting that AHQ-D scores would be statistically unrelated to grit (i.e., unwavering perseverance to achieve long-term goals, despite setbacks such as failure, disappointment, or boredom¹⁷), and we assess response bias, predicting that both social desirability response tendencies¹⁸ and beliefs in extrasensory perception¹⁹ would also be statistically unrelated to AHQ-D scores. Finally, we test whether AHQ-D scores correlate with one possible negative consequence of HF, problematic Internet use (as was the case in our prior work⁷), or with one possible positive outcome, creative achievements (as our prior work found that individuals with ADHD reported more creative achievements than those without ADHD²⁰). Our study design and these hypotheses were preregistered prior to any data collection (preregistration available at: https://aspredicted.org/ex8yr.pdf).

Methods

Scale development approach

We selected to move forward with validation of only the AHQ-D subscale (rather than also the school, hobby, screen time, and scenario HF subscales) because the AHQ-D asks about HF more generally, that is, how often HF generally occurs across different contexts. We felt that this more general phrasing would have wider applicability, rather than, for instance, asking about HF in the context of school, which applies best only to those individuals currently enrolled in high school or higher education courses. Moreover, in our prior work⁷, AHQ-D scores showed a larger difference for those with ADHD (n = 162) versus those without ADHD (n = 210; F(1,370) = 18.25, p < 0.001) compared to the other subscale and total HF scores. AHQ-D scores were also more strongly correlated (r = 0.39–0.41) with ADHD traits compared to the school and hobby subscales in our prior work⁷. Thus, the present manuscript provides psychometrics and validation of the AHQ-D in a new sample, to provide support for the AHQ-D as a valid short-form scale to quantitatively measure HF in adults.

Our next scale development step was to consider improvements to the wording of the AHQ-D items and response choices. We conducted cognitive interviews with both “think aloud” and “probing” questions, asking participants to verbalize their thought process in answering each questionnaire item, in order to understand whether participants were interpreting the AHQ-D items and response choices as we intended^21,22. We virtually interviewed 6 participants in total (2 with ADHD: 1 male (age 30), 1 female (age 27), and 4 without ADHD (1 male, 3 females, mean ± SD age = 26 ± 9 years). These interviews revealed that participants generally understood the AHQ-D questions as we intended, but that participants found the first portion of each statement to be confusing. In the previously published version, each of the 12 statements started with a phrase such as “Generally, when I am busy doing something I enjoy or something that I am very focused on…” However, interview participants indicated that the clarity of the questionnaire would be greatly improved if this statement was included as an instruction, rather than at the start of each questionnaire item. Thus, in our revised version of the AHQ-D, the statement “Generally, when I am very focused or doing something that I find especially rewarding…” is provided as a prefix to each block of six subsequent questions.

In addition, multiple interview participants found the “frequency” answer choices (i.e., six possible responses ranging from “Never” to “Daily”) to be somewhat confusing. Therefore, based on these interview results, in a subset of 196 participants, we tested one version of the AHQ-D scale in which we used six possible answer choices ranging instead from “Not at all like me” to “Very much like me”. Though the “like me” answer choices were positively correlated with the “frequency” answer choices (r(194) = 0.56, p < 0.001), the “frequency” answer choices showed stronger factor loadings, better psychometric properties, better convergent validity with the Ozel-Kizil HF scale, stronger correlations with ADHD traits, and better differentiation between those with and without ADHD (see Appendix A for details). Therefore, we decided to move forward with validating the original AHQ-D as previously published⁷, using the original “frequency” answer choices (but, as described above, revising the “Generally…” instruction to occur only twice instead of at the start of each item).

In the remainder of this manuscript, we focus on validating the AHQ-D in a new sample of adults. We conducted each of the analyses outlined for scale validation in our preregistration (https://aspredicted.org/ex8yr.pdf). Note: in our preregistration, we initially anticipated naming the updated scale the AHQ-2, rather than the AHQ-D, as we expected to make changes to the wording of the scale items and response choices; however, as the above results supported retaining our original phrasing, we also retained the original AHQ-D name.

Participants

We recruited a new sample of adults living in the United States via Prolific.co, an online platform specifically designed for recruiting research participants²³. We administered all questionnaires to participants remotely via the online platform Qualtrics. All participants were located in the United States, fluent in English, and had at least a 95% approval rating on previous Prolific study submissions. Of note, these were the only inclusion criteria; we did not systematically recruit across age, sex, or other demographic factors and therefore this sample was one of convenience and not meant to be representative of a specific geographic area.

We collected complete datasets from 401 individuals. We determined this sample size based on the recommendation of Charter et al.²⁴ to include at least 400 subjects for reliability and validity studies; in addition, we calculated that 400 participants would yield over 80% power with equivalence bounds of r = − 0.15 and r = 0.15 for the equivalence tests (for the cases in which we predicted no association between HF and a variable; see the "Statistical analyses of additional questionnaire data" section). In our preregistration, we planned to exclude any individuals who reported diagnoses of autism spectrum disorder (n = 12), bipolar disorder (n = 15 additional), obsessive compulsive disorder (n = 23 additional), schizophrenia (n = 3 additional), or borderline personality disorder (n = 1 additional). These excluded individuals responded “yes” to the questionnaire item “Have you ever been diagnosed by a doctor, psychologist, or other healthcare/medical professional with…” for the respective condition. We also planned to exclude any individuals who completed the entire survey in less than 10 min or improperly answered any of the attention checks; however, we did not need to exclude any additional individuals for these reasons.

Following these exclusions, the final sample consisted of n = 347 individuals (mean ± SD age: 33 ± 11 years; age range: 18–70 years; 47% female: 163 female, 182 male, 2 other/preferred not to answer). 87% of individuals reported their ethnicity as not Hispanic or Latino. 74% of individuals reported their race as White, 9% Black, 8% Asian, 5% Biracial, and the remaining “Other” or preferred not to answer. The cohort reported an average of 16 ± 3 years of education, and 97% reported English as their first language. Of the 347 individuals comprising the total sample, 28 participants reported a prior ADHD diagnosis by a healthcare professional (i.e., answering “yes” to the questionnaire item “Have you ever been diagnosed by a doctor, psychologist, or other healthcare/medical professional with Attention- Deficit Disorder (ADD) or Attention-Deficit Hyperactivity Disorder (ADHD)?”). The demographics of these 28 individuals who reported an ADHD diagnosis were similar to those of the larger cohort (mean age 30 ± 7 years, 89% not Hispanic or Latino, 75% White, average 16 ± 3 years of education, 100% reporting English as their first language), though the percentage of females was numerically higher in this ADHD subset (61% female: 17 female, 11 male) compared with the larger sample. Of note, in the present work we did not aim to specifically recruit individuals with ADHD; we only examine group differences between those with and without ADHD as an exploratory post hoc analysis.

Dispositional adult hyperfocus questionnaire (AHQ-D)

The AHQ-D is comprised of the “dispositional” sub-section of the full AHQ, which was first presented in our previous work⁷. The AHQ-D includes 12 items, with responses ranging from 1 (“Never”) to 6 (“Daily”). AHQ-D scores were calculated as the sum of the 12 items, with a minimum possible score of 12 (low HF) and maximum possible score of 72 (high HF). Appendix B contains a full paper-and-pencil version of the AHQ-D, as well as administration and scoring information.

Factor and item response theory analyses

We conducted all analyses using R 4.3.2²⁵ within RStudio²⁶.

Exploratory factor analysis

In our preregistration, we predicted that an exploratory factor analysis would result in high factor loadings (> 0.5) onto one general HF factor (though we opted to run an exploratory rather than confirmatory factor analysis here to avoid imposing this as a constraint and to allow for the possibility that a multi-factor solution might better fit the data). We used the psych package²⁷ to first assess the factorability of the data via Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin measure of sampling adequacy. We then conducted a parallel analysis and examined the scree plot to determine an appropriate number of factors to extract and used the fa function to run the exploratory factor analysis.

Item response theory (IRT) analysis

To further examine the AHQ-D items and response choices, we conducted an item response theory (IRT) analysis using the mirt package²⁸. IRT is based on the premise that the probability of a certain response to a questionnaire item is a function of an underlying latent dimension, denoted by theta (θ). Thus, in this case, θ represents an individual’s true, underlying general HF trait. In IRT analysis, θ is standardized to follow a normal distribution with a range from  −  3 to + 3, where 0 represents the average score²⁹. We fit the data using a graded response model, which is the model recommended for ordered polytomous response data (i.e., item responses with multiple ordered categories, such as a Likert scale)^30,31. We examined three indices to evaluate the IRT model fit: root mean square error of approximation (RMSEA), standardized root mean square residual (SRMSR), and comparative fit index (CFI), as well as one index of item fit: RMSEA S-X².

After establishing the adequacy of model and item fit, we computed item parameters. The IRT parameterization generates two metrics that assess the relationship between each item and the latent trait: slope and location parameters. Slope (also called “a” or “discrimination”) parameters indicate how well each item discriminates between different levels of the latent HF trait, as well as the strength of the relationship between each item and the latent HF trait (similar to factor loadings in the exploratory factor analysis). Slope values are on a logit scale and can range from − ∞ to + ∞ (but usually fall between 0 and 3^32,33) and indicate the strength of the relationship between the latent HF trait level and the probability of endorsing a certain response category. Items that yield higher slope values serve as better indicators of the underlying HF trait and better distinguish between individuals with different levels of the latent trait. Here we considered slope values of 0.01–0.34 “very low”, 0.34–0.64 “low”, 0.65–1.34 “moderate”, 1.35–1.69 “high”, and > 1.70 “very high”²⁹. In addition, we examined location (“b” or “difficulty”) parameters for each item, to determine if the responses covered a wide range of the latent HF trait. As the AHQ-D includes 6 possible responses, the IRT calculated 5 location parameters for each item. In the case of a graded response model, these location parameters indicate the points at which a respondent would have the same probability of endorsing any of the categories below the threshold as they would for the categories above the threshold. Thus, location parameters that cover a wide range are most desirable as they allow for greater precision in estimating one’s level of the latent trait.

Finally, we plotted item information curves. Here “information” refers to the ability of each item to accurately estimate scores on θ (the latent HF trait). That is, item-level information clarifies how well each item contributes to overall score estimation precision; higher levels of information lead to more accurate estimates of θ (the latent HF trait). In our polytomous model, the amount of information that an item contributes depends on its slope parameter (where larger slopes contribute more information) and its location parameters (where larger ranges of the location parameters indicate that an item contributes more information). We also plotted category response curves to visualize the probability of responding to specific answer choices on each AHQ item. These curves have a functional relationship with θ (the latent HF trait); that is, as θ (the latent HF trait) increases, the probability of endorsing a category (e.g., “Never”) increases at first and then decreases as responses transition to the next higher category (e.g., “1–2x every 6 months”).

Internal consistency

We used the psych package²⁷ to assess the internal consistency of the AHQ-D. We calculated Cronbach’s alpha, a metric of item inter-relatedness, or the extent to which all 12 items in the AHQ-D measure the same construct. In addition, we calculated the minimum, maximum, and mean split-half reliability; this process divides the 12 items into two random halves and compares the correlation of the results obtained from each possible split.

Additional questionnaires

ADHD traits

We first tested the correlation of AHQ-D scores with ADHD trait scores, in an effort to replicate our previous finding of a moderate association between higher HF and more ADHD traits⁷. As in our prior work, we measured ADHD traits using the Conners' Adult ADHD Rating Scale (CAARS) screening version³⁴. The CAARS Screening Version is comprised of 30 items that inquire about various ADHD traits, with responses ranging from 0 (“Not at all, never”) to 3 (“Very much, very frequently”). We calculated a raw sum score for the 30 items; possible sum scores ranged from 0 (no ADHD traits) to 90 (high ADHD traits). As recommended³⁴, we also calculated CAARS t-scores based on age and sex for 4 subscales: Inattention, Hyperactivity/Impulsivity, ADHD Symptoms, and ADHD Index. T-scores were not calculated for two participants because (as described in the "Participants" section), they preferred not to report their sex. The CAARS is a common and validated questionnaire for measuring ADHD traits. CAARS sub-scores have high internal consistency (Cronbach’s alpha 0.74–0.95)³⁵, high test–retest reliability (median correlation coefficient 0.89)³⁶, and an overall diagnostic efficiency rate of 85%³⁶.

Convergent validity

Next, we tested the correlation of AHQ-D scores with scores on several other related questionnaires to assess convergent validity, or how closely the AHQ-D is related to other tests that measure similar constructs.

Ozel-Kizil HF scale

We tested the correlation of scores on the AHQ-D with scores on the one other published quantitative HF questionnaire⁵. Prior work using the Ozel-Kizil HF scale found moderate internal consistency among a non-clinical sample of adults (Cronbach’s alpha 0.74), as well as higher HF among adults with ADHD compared to those without ADHD⁶. This scale was originally administered to participants in Turkish; though Ozel-Kizil and colleagues have published an English version of the scale⁶, we felt that this English translation included confusing wording that could be challenging for native English speakers to interpret. Thus, we re-translated the original Turkish scale to English and administered this version to the participants in the current study (see Appendix C for details). The Ozel-Kizil HF scale contains 11 items which inquire primarily about negative aspects of HF and executive dysfunction. Responses range from 1 (“Strongly disagree”) to 4 (“Strongly agree”). We calculated a sum score of the 11 items; thus, possible scores ranged from 11 (low HF) to 44 (high HF). One participant did not complete the Ozel-Kizil HF questionnaire, so n = 346 individuals for the statistical analysis of these scores.

Flow

As a second metric of convergent validity, we examined whether AHQ-D scores were associated with higher dispositional flow scores, as measured by the LONG Dispositional Flow Scale 2-General^13,14. Participants first identified an activity that typically causes them to have “peak experiences” (e.g., skiing) and responded to items asking about aspects such as their sense of control, enjoyment, and perception of time while engaging in this activity. This scale included 36 questions, with 4 items for each of 9 flow dimensions and responses ranging from 1 (“Never”) to 5 (“Always”). As recommended³⁷, we calculated the average score for each of the 9 flow dimensions, and then calculated a total flow score as the sum of these dimension scores. Total dispositional flow scores thus ranged from 9 (low dispositional flow) to 45 (high dispositional flow). Reports of internal consistency for the LONG Dispositional Flow Scale 2-General are generally high, with Cronbach’s alpha estimates ranging from 0.78 to 0.90 in healthy adults¹⁴.

Mind wandering

As a third metric of convergent validity, we examined whether higher AHQ-D scores were associated with more frequent mind wandering. We selected mind wandering as a convergent measure as we view HF to be one facet of poor attentional control and mind wandering as another, depending on an individual’s motivation towards a task⁸. We measured mind wandering using the Mind Wandering: Spontaneous scale¹⁵, which contains 4 questions that ask about unintentional mind wandering tendencies in everyday life and was found to have high internal consistency among healthy adults (Cronbach’s alpha 0.88)¹⁵. Responses were intended to range from 1 (“Almost never”) to 7 (“Almost always”); however, our survey set 0 as “Almost never.” Thus, in the present study, our possible scores ranged from 0 (low mind wandering) to 28 (high mind wandering), instead of ranging from 4 to 28.

Divergent validity

We then tested the correlation of AHQ-D scores with scores on the Grit Scale to assess divergent validity, or whether the AHQ-D is statistically unrelated to a test that measures an unrelated construct. We measured grit using the 10-item Grit Scale, available at: https://angeladuckworth.com/grit-scale/ and adapted from the original 12-item Grit Scale¹⁷, which has high internal consistency (Cronbach’s alpha 0.85). Possible responses range from 1 (“Not like me at all”) to 5 (“Very much like me”), with several reverse-coded items. We calculated the sum of the 10 item scores and divided by 10¹⁷; thus, total grit scores thus ranged from 1 (not at all gritty) to 5 (extremely gritty).

Response bias

We examined two additional metrics to assess for possible biases in the way in which participants respond to self-report questionnaires, i.e., if participants tend to be biased by what is typically perceived to be socially desirable or if they tend to respond with extreme answers to all questions regardless of topic.

Social desirability

First, we measured the relationship between AHQ-D scores and participants’ tendency to answer questions in a socially desirable (rather than a truthful) manner. We used the 13-item short form¹⁸ of the Marlowe–Crowne Social Desirability Scale³⁸. This scale has acceptable reliability (Kuder-Richardson formula 20 reliability 0.76)¹⁸ and includes 13 true (= 1) or false (= 2) statements, with 5 reverse-coded statements. Scores can thus range from 13 (low social desirability response tendency) to 26 (high social desirability response tendency).

Extrasensory perception

Second, to test whether participants with extreme AHQ-D scores tend to answer any questionnaire they encounter with extreme responses, we recorded responses on a completely unrelated questionnaire, the extrasensory perception scale¹⁹. This scale asks participants to rate their agreement with items such as “Some people have the ability to predict the future” in order to probe one’s beliefs in the existence of psychic or other extrasensory abilities. We therefore we presumed that this metric would be totally unrelated to HF tendencies. Thus, we predicted that these scores would not correlate at all with AHQ-D scores; if extrasensory perception belief scores were to correlate with AHQ-D scores, it might suggest that participants are biased in the way they answer any questionnaire presented to them (e.g., incorrectly always selecting the weakest response choice in order to complete the items as quickly as possible). This scale has high internal consistency (Cronbach’s alpha 0.91) and includes 14 items that measure beliefs regarding the existence of psychic or extrasensory abilities. Responses range from 1 (“Strongly disagree”) to 5 (“Strongly agree”), with 4 reverse-coded items. Possible total scores thus range from 14 to 70.

Internet addiction and creative achievement

Finally, we tested whether AHQ-D scores correlated with one potential negative consequence of HF: problematic Internet use (as in our prior work⁷), and one potential positive consequence of HF: creative achievements.

Internet addiction

We measured problematic Internet use with the Internet Addiction Test³⁹. “Internet use” was defined broadly, including use of any device with Internet access to engage in school, work, online shopping, social media, online dating, watching media online, online gaming, or miscellaneous “surfing the web.” The Internet Addiction Test includes 20 items that ask about frequency of negative consequences of Internet use (e.g., loss of sleep and neglecting other responsibilities). Responses range from 0 (“Not applicable”) to 5 (“Always”); total scores range from 0 (“healthy” Internet usage) to 100 (“severe dependence upon the Internet”).

Creative achievements

Participants completed the Creative Achievement Questionnaire⁴⁰, which measures creativity in 10 domains (visual arts, music, dance, architectural design, creative writing, humor, invention, scientific discovery, theater/film, and culinary arts) and has high test–retest reliability (r = 0.81) and internal consistency (Cronbach’s alpha 0.96)⁴⁰. Each domain includes 7 items that ask about domain-specific activities and accomplishments (e.g., “I have choreographed dance professionally”). Various items ask participants to indicate the number of times that a statement applies to them (e.g., “I have received a grant to pursue my work in science or medicine”); thus, the minimum possible score is 0, but there is no maximum possible score. We examined three different scores: raw sum score (n = 347), raw sum score excluding those with a total score of 0 (n = 269), and (due to the non-normality of the sum score distribution and following prior work^20,41) log-transformed total scores excluding those with a total score of 0 (also n = 269).

Statistical analyses of additional questionnaire data

As outlined in our preregistration, we calculated the Pearson correlation between AHQ-D scores and scores for each questionnaire described above. The AHQ-D sum score (as well as several other questionnaire scores) did not meet the normality assumption (Shapiro test p < 0.05); however, rerunning these instead as nonparametric Spearman correlations did not change the interpretation of any correlation direction, strength, or statistical significance. Therefore, we report Pearson correlations throughout the manuscript.

In the several cases for which we predicted no association between HF and a variable (i.e., grit, social desirability, and extrasensory perception), we also conducted an equivalence test using the TOSTER package⁴². We set equivalence bounds of r = − 0.15 and r = 0.15 (i.e., indicating that weaker correlation strength than 0.15 is equivalent to the absence of a worthwhile effect). We then used the TOSTr function to test whether the correlation between AHQ-D and each of these three variables fell within the equivalence bounds (which would provide further confirmation that that two variables are statistically unrelated).

Ethics approval and consent to participate

The University of Michigan Institutional Review Board approved all study procedures, and written informed consent was obtained from all participants.

Results

Scale properties for the AHQ-D

Exploratory factor analysis (EFA)

Both Bartlett’s test of sphericity (χ² = 2757.28; p < 0.001) and the Kaiser–Meyer–Olkin measure of sampling adequacy (overall MSA = 0.92) suggested that an exploratory factor analysis would be appropriate for these data. The scree plot and parallel analysis supported a one-factor solution; the eigenvalue was > 1 for only the first factor, and the ratio of the first to the second eigenvalue was 7.00. The one-factor solution explained 54.3% of the variance in HF responses, and each of the questions loaded highly onto this one factor (factor loadings 0.57–0.81; Table 1). Thus, as hypothesized, the data supported the unidimensionality of HF, i.e., one general dispositional HF factor.

Table 1.

Exploratory factor analysis (EFA) and item response theory (IRT) results.

AHQ-D Scale Item	EFA factor loading	RMSEA S-X2	a	b1	b2	b3	b4	b5
1. I tend to completely lose track of the time	0.81	0.021	2.70	− 1.45	− 0.69	0.16	0.70	1.69
2. I do not notice the world around me, and I won’t realize if someone calls my name or if my phone buzzes	0.65	0.023	1.64	− 0.62	0.12	0.70	1.49	2.67
3. I might accidentally miss meals, stay up all night, or keep doing the activity until I absolutely must get up to go to the bathroom	0.75	0.025	2.10	− 0.84	− 0.21	0.44	1.01	1.87
4. I find it very difficult to quit and move on to doing something else, even if I have a lot of other important things I should be doing instead	0.77	0.000	2.46	− 1.15	− 0.46	0.36	0.99	1.89
5. I can feel totally captivated by or “hooked” on the activity	0.73	0.028	2.19	− 1.97	− 1.09	− 0.14	0.59	1.62
6. I sometimes focus for far too long on a small detail of the task and avoid other important parts	0.79	0.010	2.55	− 1.19	− 0.43	0.34	1.08	1.93
7. I can be unsure of what time of day it is or how much time has passed since I started the activity	0.73	0.014	2.04	− 1.09	− 0.29	0.43	1.10	2.16
8. I don’t react to any distractions (e.g., if someone talks to me)	0.57	0.015	1.27	− 0.79	0.27	1.10	2.12	3.40
9. I forget to attend to my personal needs (e.g., I forget to sleep or eat, or I wait until the last minute to go to the bathroom)	0.75	0.000	2.22	− 0.68	− 0.12	0.46	1.04	2.05
10. I feel like I can’t stop doing the activity, even if I have other more important responsibilities	0.79	0.005	2.61	− 1.04	− 0.31	0.41	1.13	2.04
11. I can feel completely engrossed or fixated with the activity	0.72	0.000	2.19	− 2.04	− 0.96	− 0.13	0.50	1.67
12. I can get “stuck” on little details that keep me from finishing other important parts of the task	0.76	0.029	2.36	− 1.17	− 0.50	0.23	1.05	2.10

EFA loading indicates the exploratory factor analysis loading of each item onto the single HF factor. RMSEA S-X² assesses the degree of item fit (“good” fit: < 0.06).

a indicates the IRT slope or discrimination parameters, and b1–b5 indicate the IRT location or difficulty parameters.

Item response theory (IRT) analysis

The IRT model yielded the following fit indices: RMSEA = 0.135 (“good” fit ≤ 0.06), SRMSR = 0.066 (“good” fit ≤ 0.08), and CFI = 0.950 (“good” fit ≥ 0.95). Thus, two of the three fit parameters (SRMSR, CFI) indicated a “good” fit of the IRT model. While the RMSEA metric did not indicate a “good” fit, prior simulation work⁴³ demonstrates how high factor loadings (as was the case here) can bias RMSEA to be higher than the traditional cut-off of ≤ 0.06. The RMSEA S-X² metric of item fit was < 0.06 for all items (Table 1), suggesting that all items had adequate fit with the model.™

After establishing this adequacy of model and item fit, we then computed item parameters, the first of which was slope (also called a or discrimination parameters). Slopes ranged from 1.27 to 2.70 (Table 1), with the first AHQ-D item being the most discriminating (slope = 2.70), and the eighth item being the least discriminating (slope = 1.27). Higher slope values indicate greater discriminatory power, and in this case, all items exhibited “moderate” (item 8), “high” (item 2) or “very high” (remaining 10 items) discrimination²⁹. This suggests that each item effectively distinguishes between individuals with high and low levels of the latent HF trait. Secondly, we examined location parameters (b1–b5 in Table 1), which revealed that the responses covered a wide range of the latent HF trait. A broad coverage is desirable as it enhances precision in estimating one’s level of the latent trait. Both the slope and location parameters align with the item information curves (Fig. 1), which illustrate how well each item contributes to overall score estimation precision. Larger slopes and wider ranges of location parameters contribute more information. Here we can see that item 1 was the most informative item, while items 2 and 8 were the least informative.

Figure 1.

Item Information Curve from the IRT Analysis. θ (the latent HF trait) is depicted on the x-axis, and information is depicted on the y-axis. Higher values on the y-axis indicate that the item is more informative for estimating a person's latent HF at that trait level. In other words, a higher information curve signifies greater precision in the item’s ability to differentiate between individuals with varying levels of latent HF.

Figure 2 presents category response curves to visualize the probability of responding to specific answer choices on each AHQ item. The category response curves were largely symmetrical, indicating that the six response choices covered a wide range of the latent HF trait θ (and thus appropriately covered individuals with varying levels of HF), and that response choices were ordered correctly.

Figure 2.

Category Response Curve from IRT Analysis. θ (latent HF) is depicted on the x-axis, and the probability of endorsing a specific response category is depicted on the y-axis. These curves indicate that the response categories covered a wide range of θ, and that the response choices were ordered correctly. As θ (latent HF) increases, the probability of endorsing a category (e.g., “1–2x every 6 months”) increases at first and then decreases as responses transition to the next higher category (e.g., “1–2x per month”).

Internal consistency

The AHQ-D yielded high internal consistency: Cronbach’s alpha 0.93 and split-half reliability minimum 0.88, mean 0.93, and maximum 0.97.

Descriptive statistics

Table 2 presents descriptive statistics (i.e., mean, standard deviation, minimum score, maximum score, skewness, and kurtosis) for AHQ-D scores, as well as the additional scale metrics, across the whole sample. Appendix D contains a correlation matrix that depicts the correlations between each of the outcome metrics.

Table 2.

Descriptive statistics for AHQ-D scores and additional scale metrics.

Outcome metric	Mean	SD	Min	Max	Skewness	Kurtosis
AHQ-D	36.90	13.54	12	72	0.30	− 0.79
CAARS (ADHD Traits)	27.24	15.33	0	82	0.52	− 0.20
Ozel-Kizil HF	25.59	5.76	11	42	0.02	− 0.05
Flow	36.55	4.38	19.75	45	− 0.27	0.27
Mind wandering	13.86	7.21	0	28	0.01	− 0.91
Grit	3.25	0.72	1.4	5	− 0.07	− 0.25
Social desirability	18.51	3.07	13	26	0.18	− 0.56
Extrasensory Perception	31.06	13.02	14	68	0.50	− 0.70
Internet addiction	30.87	14.86	0	82	0.78	0.82
Creative Achievement	1.96	1.14	0	5.43	0.14	− 0.22

These descriptive statistics were calculated for each of the respective questionnaires across the entire (n = 347) sample—with the exception of the Ozel-Kizil HF scale which one participant did not complete (thus n = 346 for this metric) and creative achievement scores which are the log-transformed total scores excluding those with a total score of 0 (n = 269). SD, standard deviation; AHQ-D, dispositional adult hyperfocus questionnaire; CAARS, Conners’ Adult ADHD Rating Scale (total raw score); HF, hyperfocus.

Relationship of HF with ADHD traits

As predicted, higher AHQ-D scores (i.e., higher HF) correlated with higher CAARS scores (i.e., higher ADHD traits). This was the case for total raw CAARS score (r(345) = 0.53, p < 0.001; Fig. 3A) and for the four CAARS t-score subscores calculated based on age and sex: Inattention (r(343) = 0.49, p < 0.001), Hyperactivity/Impulsivity (r(343) = 0.43, p < 0.001), ADHD Symptoms (r(343) = 0.52, p < 0.001), and ADHD Index (r(343) = 0.51, p < 0.001). A post hoc two-sample t-test revealed that AHQ-D scores were higher for those who self-reported an ADHD diagnosis by a healthcare professional (mean 45.64 ± 12.46) compared to those without ADHD (mean 36.02 ± 13.34; t(32.7) = -3.89, p < 0.001; Fig. 3A). Note that this two-sample t-test was a post hoc exploratory analysis, intended to replicate our previous finding of higher HF in those with self-reported ADHD⁷. In the present work, we did not recruit equal cohorts of participants with and without ADHD; thus, given the unequal group sizes here (n = 28 with ADHD versus n = 317 without ADHD), this result should be interpreted with caution.

Figure 3.

Correlations of AHQ-D Scores with Additional Scale Metrics. The bottom righthand corner of each plot indicates the Pearson correlation coefficient, degrees of freedom, and p-value significance. Statistical significance of each correlation is indicated as: *p < 0.05, **p < 0.01, ***p < 0.001. n = 347, with the exception of the Ozel-Kizil HF scale which one participant did not complete (thus n = 346 for this metric) and creative achievement scores which are the log-transformed total scores excluding those with a total score of 0 (n = 269). AHQ-D, dispositional adult hyperfocus questionnaire; CAARS, Conners’ Adult ADHD Rating Scale (total raw score); HF, hyperfocus.

As ADHD is more likely to occur in males compared with females⁴⁴, we conducted an additional post hoc exploratory analysis examining whether AHQ-D scores differ based on sex. On average, males had slightly higher AHQ-D scores (mean 38.27 ± 13.44) compared with females (mean 35.13 ± 13.41; t(339.02) = -3.89, p = 0.031).

Convergent validity

The AHQ-D displayed the hypothesized associations with each of the convergent validity metrics (Fig. 3B). Higher AHQ-D scores (higher HF) correlated strongly with higher HF measured using the Ozel-Kizil HF scale (r(344) = 0.69, p < 0.001). There was a weak positive correlation between AHQ-D scores and flow (r(345) = 0.12, p = 0.028) and a moderate positive correlation between AHQ-D scores and mind wandering (r(345) = 0.39, p < 0.001).

Divergent validity

Though we hypothesized no relationship between AHQ-D scores and grit, the data revealed a weak negative correlation between HF and grit (r(345) = − 0.29, p < 0.001; Fig. 3C). The TOST test similarly revealed a non-significant result (p = 0.997; i.e., indicating that the 0.29 correlation did not fall within the equivalence bounds). Thus, we can conclude that higher HF is significantly (but weakly) associated with lower grit.

Response bias

As predicted, there was no relationship between beliefs in extrasensory perception and AHQ-D scores (r(345) = 0.01, p = 0.834; Fig. 3C). Further confirming this, the TOST test was statistically significant (p = 0.005), indicating no meaningful relationship between extrasensory perception and HF. Contrary to our hypothesis predicting no association between HF and social desirability, we did find a weak negative correlation in which higher HF correlated with lower social desirability response tendency (r(345) =  − 0.24, p < 0.001; Fig. 3C). The TOST test similarly revealed a non-significant result (p = 0.959), indicating that the − 0.24 correlation fell outside of the equivalence bounds and thus that higher HF is significantly (but weakly) associated with lower social desirability response tendency.

Internet addiction and creative achievement

As hypothesized, higher AHQ-D scores were moderately correlated with higher Internet addiction scores (r(345) = 0.44, p < 0.001; Fig. 3D). Also as hypothesized, higher AHQ-D scores were weakly correlated with more creative achievements. This was the case for raw Creative Achievement Questionnaire scores (r(345) = 0.14, p = 0.009), as well as scores excluding sum scores of 0 (r(267) = 0.13, p = 0.035) and the log-transform of creative achievement scores (r(267) = 0.17, p = 0.004; Fig. 3D).

Discussion

Results overview

In the present study we evaluated whether the Dispositional Adult Hyperfocus Questionnaire (AHQ-D) could serve as a validated shortened version of the full Adult Hyperfocus Questionnaire (AHQ; originally published in 2019⁷), that appropriately and efficiently measures HF in adults. In a new sample of adults (n = 347), we found the AHQ-D to be a valid scale for measuring HF based on: (1) the generation of high factor loadings on a unitary HF construct; (2) an IRT analysis that yielded high item-level slopes (indicating the utility of each item for differentiating between persons with high and low levels of the latent HF trait) and a wide range of location parameters (which allows for greater precision in estimating one’s level of the latent HF trait); and (3) high internal consistency metrics. Our results also revealed the hypothesized relationship between higher HF scores and more ADHD traits, as well as the hypothesized convergent validity and a general lack of response bias. We did identify a weak negative correlation between AHQ-D scores and grit (our metric of divergent validity for which we anticipated no statistical relationship with HF scores); we discuss possible reasons for this association below. Finally, as an exploratory analysis, we identified potential positive and negative consequences of HF. Taken together, we suggest that the AHQ-D is a valid metric for quantitatively assessing HF in adults.

Scale properties

In investigating the AHQ-D scale properties via exploratory factor analysis, we found that one latent variable could be extracted amongst the 12 AHQ-D items. That is, all scale items satisfactorily loaded onto the same factor (factor loadings 0.57–0.81), further suggesting that the AHQ-D measures a unitary HF construct. This is in line with both our preregistered hypothesis (predicting a unitary HF construct), as well as our prior work⁷ which considered all 66 items from the full AHQ in one exploratory factor analysis and found that the majority of the dispositional HF items (11 of 12 items) all loaded onto a single dispositional HF factor. Next, we utilized IRT analysis to evaluate the usefulness of each item within the AHQ-D scale and determined that all items should be retained. IRT slope parameters yielded “moderate” (item 8), “high” (item 2) and “very high” (remaining 10 items) discrimination²⁹, indicating the utility of each item for differentiating between individuals with high and low levels of the latent HF trait. Location parameters indicated that the responses covered a wide range of the latent HF trait (which is desirable). While the item information curves suggested that items 2 and 8 provided comparatively less information (i.e., less score estimation precision), these items still had high factor loadings (0.65 and 0.57, respectively) and therefore we decided to retain these items and thus support the use of the dispositional subscale in its original form (with its original item wording and response choices). Finally, our assessment of the AHQ-D revealed high Cronbach’s alpha and mean split-half reliability (both 0.93). This is in line with two independent studies which similarly reported Cronbach’s alpha values of 0.94¹¹ and 0.90⁹ for the AHQ-D. In comparison, the Ozel-Kizil scale reported a comparatively lower Cronbach’s alpha of 0.74 in a sample of 207 adults⁶.

Relationship of HF with ADHD traits

As hypothesized, we found that higher AHQ-D scores (higher HF) correlated with higher CAARS ADHD trait scores; this was the case for both the total raw CAARS score (r = 0.53), as well as the four CAARS t-score subscores (r = 0.43–0.52). This is in line with our prior work in which we found significant but comparatively weaker correlations (r = 0.39–0.41) between CAARS t-score subscores and the dispositional HF subscale⁷. This finding also corresponds with three recent studies by other groups in general population samples (n = 84¹¹, n = 380⁹, and n = 694¹²) which similarly found significant positive correlations (r = 0.70, 0.47, and 0.31, respectively) between AHQ-D scores and ADHD traits measured using the Adult ADHD Self Report scale (ASRS)⁴⁵. Furthermore, this result is in line with an additional study which found that ADHD traits measured using the ADHD Rating Scale (ARS)⁴⁶ were weakly positively correlated with HF frequency (i.e., participants’ response to “how often do you experience these periods” among the 929 of 1124 participants who reported HF occurrences)⁴⁷. Taken together, the present results provide new empirical evidence to support the documented positive relationship between ADHD traits and dispositional HF measured using the AHQ-D scale.

Additionally, our exploratory post hoc analysis revealed that AHQ-D scores were significantly higher for those who self-reported an ADHD diagnosis compared to those without ADHD. This group difference replicates our prior report of dispositional HF differences by ADHD diagnosis⁷ in a new cohort. However, it is important to acknowledge that (unlike our prior work⁷) here we did not specifically recruit individuals with ADHD and therefore this group difference should be interpreted with caution. Limited prior work^10–12,47 has similarly compared HF by ADHD diagnosis (though mostly in post hoc analyses and/or relatively small samples). One of these studies¹¹ found that 41 college students with high ADHD traits (ASRS score ≥ 14) had higher AHQ-D scores compared to 43 students with low ADHD traits (ASRS score < 14); the same study also found that the 20 students in this sample who self-reported a prior ADHD diagnosis also had significantly higher AHQ-D scores compared with the 64 who did not. Similarly, the second¹² and third⁷ studies reported higher AHQ-D scores for those who self-reported an ADHD diagnosis (n = 20 and n = 122, respectively) compared with those who did not (n = 674 and n = 99). The fourth study⁴⁷ found no difference in HF occurrence, frequency, duration, or pervasiveness between 78 individuals with ADHD (confirmed by clinical psychiatric interviews) and 78 matched controls, concluding that HF might not be specific to ADHD. However, this study used a new 4-question scale (e.g., asking “yes” or “no” for whether individuals ever experience HF). Therefore, while interesting, this result is not directly comparable to our analysis of AHQ-D scores. Thus, while our present work, prior study⁷, and the work by Grotewiel et al.¹¹ each suggest group differences in AHQ-D scores based on ADHD diagnosis, future studies should include targeted recruitment of large samples of those with and without ADHD to more definitively replicate this finding. Future work might also consider targeted recruitment of individuals with other demographic factors and diagnoses (e.g., autism¹⁰) in order to better understand what other individual differences might contribute to HF tendencies.

Convergent validity

We first assessed the convergent validity of the AHQ-D by evaluating its association with the Ozel-Kizil HF scale⁵. We found a strong positive correlation (r = 0.69) between HF scores on the AHQ-D and our English translation of the Ozel-Kizil HF scale, suggesting that the AHQ-D measures a similar HF construct to that of the Ozel-Kizil scale. Similarly, another independent study⁹ found a strong positive correlation (r = 0.70) between AHQ-D scores and the authors’ 10-question adaptation of the Ozel-Kizil HF scale. Previously, we critiqued the Ozel-Kizil scale’s 11 items for focusing primarily on possible negative consequences of HF (e.g., neglecting others, feeling pain in the body, and disrupting relationships)⁷, whereas we aimed to word the AHQ-D items in a more neutral manner that focused on the subjective experience of HF, rather than possible consequences of HF. Moreover, we also previously criticized the Ozel-Kizil scale for including some items that might better be classified as executive function deficits (e.g., failing to complete work that one has started) as opposed to HF⁷. Therefore, though we (and others⁹) addressed the English language issue by using our own translation of the scale, we still suggest that our AHQ-D questionnaire provides a more accurate representation of dispositional HF tendencies based on our working definition of HF. However, we found it both interesting and reassuring that we (and others⁹) identified such a strong correlation between the AHQ-D and the Ozel-Kizil HF scale; thus, it is reasonable to conclude that the Ozel-Kizil HF scale also provides valuable insights into HF in adult populations.

We then assessed the convergent validity of the AHQ-D by evaluating its association with a metric of flow^13,14. We identified a weak positive correlation (r = 0.12) between AHQ-D scores and flow scores. This relationship was somewhat weaker but in the same direction as the relationship we identified between dispositional HF and flow in our prior work (r = 0.26)⁷. This result suggests that, as anticipated, HF and flow are distinct but related constructs. That is, both involve intense focus on an activity, but flow is typically considered to be more of a productive or “peak”/“optimal” experience^16,48, whereas HF may have positive (e.g., completing a creative project) or negative/neutral outcomes (e.g., watching many hours of television). Moreover, as we discussed previously⁷, HF could be a type of “deep” flow, which is a more intense experience that involves detachment from one’s environment^11,49; however, the flow questionnaire^13,14 primarily asks about aspects of “shallow” flow, and therefore it likely follows these flow scores would be positively related to AHQ-D scores, but exhibit a weak rather than strong association. This fits with another study¹¹ that used the same dispositional flow scale^13,14, but separately examined correlations of the AHQ-D with each of the 9 flow dimensions (instead of with total flow score). They identified a positive correlation (r = 0.50) between AHQ-D scores and the “time transformation” flow dimension, in addition to negative correlations between AHQ-D scores and the “goals”, “feedback”, “concentration”, and “control” flow dimensions. The authors suggest that these results support the notion of HF representing a type of “deep” flow, involving detachment from the environment in a more extreme manner than during “shallow” flow experiences. Taken together, it is likely that HF and these flow scores are related in some aspects (e.g., both including losing one’s sense of time), but different in other aspects (e.g., “shallow” flow requires clear goals and a sense of control). Thus, the weak positive correlation between the AHQ-D and total flow scores identified here is likely reasonable; however, future studies might also explore HF relationships with reports of both “shallow” and “deep” flow.

Lastly, we examined the convergent validity of the AHQ-D by examining its association with spontaneous mind wandering¹⁵. We found a moderate (r = 0.39) correlation between AHQ-D and mind wandering scores, indicating that higher HF is associated with more mind wandering. This finding appears to be in alignment with a growing body of literature suggesting that individuals with ADHD experience elevated levels of spontaneous mind wandering^50–54. We suggest that this positive association between HF and spontaneous mind wandering provides evidence of convergent validity for the AHQ-D, as it could be that individuals with poor attentional control (e.g., those with ADHD) may be predisposed to mind wandering or HF, depending on their motivational state and engagement in a task. Moreover, as discussed in our prior work⁸ (where we similarly reported a positive relationship between HF and mind wandering), it could be that both HF and mind wandering involve perceptual decoupling, or reduced processing of sensory inputs in one’s environment^55,56, thereby protecting the internal train of thought from external disruptions⁵⁷ and allowing one to fixate on their HF task, and/or mind wander.

In the consideration of HF and mind wandering, it is also important to acknowledge another dimension of mind wandering termed deliberate mind wandering, which occurs intentionally. One study reported an association between spontaneous mind wandering and inattentive, hyperactive, and impulsive ADHD traits, but no relationship between deliberate mind wandering and any ADHD traits⁵⁸. From a mechanistic perspective, Bozhilova and colleagues⁵⁹, proposed the mind wandering hypothesis, which contends that altered deactivation of the default mode network (DMN), as well as dysfunctional interaction of the DMN with the executive control network, leads to heightened spontaneous mind wandering in individuals with ADHD. Specifically, Bozhilova and colleagues⁵⁹ point to evidence that those with ADHD exhibit reduced deactivation of the DMN during tasks requiring sustained attention, which may lead to high levels of internal distractibility and, resultingly, spontaneous mind wandering. With this literature in mind, we only expected the AHQ-D to be associated with spontaneous mind wandering (and not deliberate mind wandering). However, future work might consider including a measure of deliberate mind wandering in order to specifically test this hypothesis.

Divergent validity

We examined the relationship of AHQ-D scores with grit¹⁷ as a measure of divergent validity. Contrary to our hypothesis of no statistical relationship between AHQ-D and grit scores, we identified a weak negative correlation between HF and grit, in which those with higher dispositional HF scores showed lower grit. Notably, this correlation (r = − 0.29) was substantially weaker than that of the primary convergent validity metric (HF measured using the Ozel-Kizil scale, r = 0.69), which we suggest provides evidence of the expected divergent validity. Moreover, this finding is perhaps logical, in that grit involves perseverance to achieve long-term goals, including “maintaining effort and interest over years despite failure, adversity, and plateaus in progress…whereas disappointment or boredom signals to others that it is time to change trajectory, the gritty individual stays the course”¹⁷. This is in direct contrast to HF, which entails engaging in activities for shorter periods of time (days, weeks, months) rather than years, and HF is related to intense focus on activities that one finds intrinsically motivating and engaging, rather than tedious or non-rewarding steps required to achieve long-term goals. Therefore, it follows that those who report higher levels of dispositional HF would report lower levels of grit; those with higher HF might find it more difficult to stick to something if they do not find it to be engaging. Thus, it was perhaps incorrect to hypothesize that grit would be statistically unrelated to AHQ-D scores, and instead it is likely logical that there would be a weak negative correlation between these two metrics.

Response bias

We utilized a scale probing beliefs in extrasensory perception¹⁹ to help determine whether individuals who score high on the AHQ-D tend to also obtain extreme scores on a metric totally unrelated to attention. As predicted, there was no statistical relationship between beliefs in extrasensory perception and AHQ-D scores (r = 0.01), suggesting that individuals with high self-reported HF tendencies do not always simply answer any survey question using the most extreme answer options. We also predicted that there would be no statistical relationship between AHQ-D scores and tendency to answer questions in a socially desirable (rather than an honest) manner, as measured by the Marlowe–Crowne Social Desirability Scale³⁸, which asks for self-report on items like “I am sometimes irritated by people who ask favors of me.” Contrary to our hypothesis, we did find a weak negative correlation in which higher HF correlated with lower social desirability response tendency (r = − .24). Though we aimed to word the AHQ-D items in a neutral manner, it could still be that HF is typically thought of in a negative context (e.g., resulting in neglecting oneself, their relationships, or their obligations⁶). Those who are more concerned with what others think of them (i.e., those with a higher social desirability response tendency) might then report less HF, if they view HF as primarily negative. We do not think that this weak correlation of AHQ-D scores with social desirability impairs the validity of the AHQ-D; however, future research might consider further exploration of positive versus negative perceptions of HF, and how such perceptions interact with quantitative HF scores.

Internet addiction and creative achievements

Lastly, as an exploratory aim, we examined whether AHQ-D scores associated with one “negative” (i.e., Internet addiction) and one “positive” (i.e., creative achievements) quality. We replicated our previous⁷ finding that higher HF was associated with higher Internet addiction scores; the association in the present work (r = 0.44) was stronger than the association reported in our prior work (r = 0.26). In this manner, high HF tendencies could be considered problematic for some individuals, particularly those with executive control difficulties such as persons with ADHD. However, some individuals clearly report positive outcomes of HF; for instance, short-answer responses in our prior work⁷ included descriptions of high levels of productivity and achievement resulting from HF episodes (e.g., completing art projects, creative writing, coding tasks, or complicated duties for one’s job). Moreover, our prior work identified more real-world creative achievements (as measured by the Creative Achievement Questionnaire⁴⁰) for adults with ADHD compared to those without ADHD²⁰. Past work has not investigated how HF tendencies might relate to real-world creative achievements. Here we found a weak positive correlation (r = 0.17) between creative achievements and AHQ-D scores. Thus, as AHQ-D scores were more strongly associated with Internet addiction than creative achievements, it could be that high HF tendencies might tend to be more problematic than productive. However, further work utilizing both quantitative and qualitative methods to provide additional context to individuals’ experiences is needed to more fully understand the qualities that might make an individual more likely to have negative versus positive HF outcomes.

Limitations

There are several limitations to this work. In the present study, we aimed to validate the AHQ-D in a large sample of adults; we did not specifically recruit individuals with ADHD for this cohort. Thus, the post hoc exploratory analysis reporting on AHQ-D scores in those with and without self-reported ADHD should be interpreted with caution, as we compare quite unequal group sizes (n = 28 with ADHD versus n = 319 without ADHD). The purpose of this post hoc exploratory analysis was solely to replicate our prior results which reported higher HF in those with ADHD versus those without⁷. Our prior work administered the full 66-item AHQ, whereas in the present work we aimed to demonstrate higher HF in those with ADHD using only the 12-item AHQ-D. Future studies might consider recruiting a balanced sample of those with and without ADHD and utilizing more detailed clinical measures (e.g., semi-structured interviews and other ADHD rating scales in addition to the CAARS) to more fully explore how HF interacts with ADHD diagnosis and traits. In addition, in the current work we were limited to collecting a convenience sample based on volunteers who met our enrollment criteria and were available to participate on the Prolific platform. Future studies might perform more systematic sampling across age, sex, and other demographic factors in order to collect a fully representative sample of adults living in the United States. Moreover, future work should recruit larger samples powered appropriately to conduct measurement invariance testing across sex and age groups. In the present work, we did not perform measurement invariance testing and therefore it is unknown whether the AHQ-D items measure the same construct equally across males versus females or across different ages. In addition, as present work validates the AHQ-D only in an adult (18 + years) cohort, future studies might consider further scale development efforts to adapt the AHQ-D items for pediatric populations or their caregivers. Finally, future studies might consider replicating the present results in a new adult cohort and using confirmatory factor analysis to validate the unidimensionality of the AHQ-D.

Conclusions

Together, this work supports the AHQ-D as a valid quantitative measure of dispositional HF in adults. We present strong scale metrics, the expected convergent validity, and a general lack of response bias, in addition to replicating our previous association of higher HF with ADHD traits and diagnosis. We therefore suggest that the AHQ-D represents a valid means for researchers to measure HF in the general adult population.

Author contributions

K.H. conducted all statistical analyses, created all figures and supplemental material, and wrote the first draft of the manuscript. J.O. collected all questionnaire data via Prolific and contributed to writing the Discussion. Q.T. conducted all cognitive interviews. H.H. contributed to data interpretation and writing the Introduction and Discussion. T.A. consulted on statistical design and contributed to manuscript editing. P.S. led the design of the project and interpretation of the results. All authors participated in initial experimental design, formulation of hypotheses, and revision of the manuscript.

Funding

During completion of this work K.H. was supported by National Institute on Aging fellowship K00 AG068440.

Data availability

The dataset analyzed in the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-70028-y.