Introduction

Human social interactions require the ability to engage effectively with both the environment and other individuals1. To develop this aptitude, individuals must interpret and understand situations, as well as their own and other’s actions2,3,4,5 Interpersonal differences in decoding actions, which involve synthesis of movement kinematics and setting functional goals, could find roots in the theory of action identification6,7,8.

This theory emphasizes the integration of motor and cognitive aspects in the formulation of action identification, based on the level of personal agency. Indeed, it has been postulated that individuals differ in their tendencies to understand and describe actions. Someone focuses on the details or mechanisms to describe the action (i.e., the “how”), concentrating on the modalities of execution (e.g., how to drive a car), while others emphasize the consequences (i.e., the “why”) focusing on the broader implications (e.g., drive a car to travel). The former perspective defines concrete, low-level characteristics of actions, resulting in overt description, while the latter relies on abstract, high-level characteristics from covert information. Each action could be described from both perspectives, and this dualism could coexist in every individual, varying dynamically across situations, with one trait tenting to emerge.

The modality by which humans could capture information from the external world is associated with different hierarchically organized levels of action abstraction. These levels are influenced by previous individual experiences and can be modulated by the ambiguity of context, beyond the difficulty of execution6. Therefore, an abstract level of action identification might enhance performance, which can be captured as an automated gesture through full knowledge and awareness of the action itself9, and is linked to interoceptive experience10. In contrast, when the experience is poor, more details are required to identify an action, resulting in a lower concrete level of detection, relying on sensorimotor experience11.

Although the first formulation of the Action Identification Theory is rather outdated, its concept has evolved over the years. In particular, it has been integrated with the concept of levels of abstraction12, and is now a relevant issue in both social and personal contexts, offering valuable insight into diverse aspects of human behavior13,14. Current evidence suggests that, in our complex model of society, feelings of autonomy influence the level of action identification. When individuals feel empowered and autonomous, they tend to attribute more abstract meaning to their activities, acting in a purpose-oriented manner that promote greater efficiency. Conversely, reduced autonomy can lead to more concrete action identification, diminishing the sense of responsibility and affecting motivation, self-regulation, and well-being12.

Furthermore, individual levels of action identification can be altered by some psychological disorders. For example, individual with depression or anxiety tend to assign a more abstract identifications to negative than positive ones, with impacts on their emotional responses, leading to repetitive thought patterns, and hinders problem-solving abilities15. Evidence also suggests that chronic pain may influence the level of action abstraction. Persistent symptoms could interfere with daily activities, leading individuals to perceive their actions at lower, less meaningful levels, thereby affecting their global perception of life16.

Starting from the Action Identification Theory, Vallacher developed the Behavior Identification Form (BIF)7 as an instrument to measure the personal construct of action identification. This outcome measure consists of 25 items that propose fairly common activities to which meaning can be attributed by choosing an alternative within a dual response option, one describing an interpretation of actions based on motor aspects (low-level) and the other based on more cognitive aspects (high-level). The subjects’ level of action identification is defined by the number of high-level choices. The overall score, from 0 to 25, is used as an indicator of the level of action identification, associating higher scores with preferences for higher-level action identifications and, consequently, higher-level construal mindset17.

The use of BIF to assess principles of Action Identification Theory, in which mental representations and action are strictly related to each other, assumes relevance to deepen the dynamics of human emotional-motor-cognitive behaviors immersed in different social contexts9.

Indeed, this measure may allow assessing the individual’s tendency to read the reality, as in a Turkish study in which BIF measured the construal level18 defined as the psychologically distance towards an action. Currently, there is no validated Italian version of the BIF. Introducing this instrument into the Italian context would address an existing gap by providing a culturally appropriate tool. Once its psychometric properties are established, the Italian version of the BIF could be used in both research and clinical settings involving Italian populations.

On these bases, the main aim of this study was to validate the Italian version of the BIF and assess its psychometric proprieties, in the domains of validity and reliability, in a sample of healthy adults. Secondary goal was to also explore the technical characteristics of this instrument.

Material and methods

This cross-sectional study was developed from the collaboration between the University of Genoa, the University of Udine and the IRCCS E. Medea. The first stage of the project was dedicated to the translation in Italian language of the BIF, the second one to the validation process and analysis of measurement properties of the instrument.

The second stage of this study was designed following the Consensus-based Standards for the selection of health Measurement Instruments (COSMIN) guidelines and the full reporting adhered to its format19. The study was conducted in accordance with legal and ethical international standards outlined in the updated Declaration of Helsinki20 after requesting approval from the local Committee of the University of Genoa (CERA, Unige, approval 2024/80) and obtaining written informed consent from all participants who confirmed their agreement by checking the appropriate box on the first page of the Google Form online material before proceeding with the BIF items.

Instrument and translation process

The BIF is a self-administered questionnaire consisting in 25 items investigating the level of identification of familiar, common actions according to the Action Identification Theory. Participants were asked to choose between one of two alternative descriptors, classified as low or high level, of a defined action. High level responses contributed to generate the total score of the questionnaire, ranging from 0 to 25, and defining the level of action identification characterizing each subject.

Italian translation was conducted following the “Translation and Cultural Adaptation of Patient Reported Outcomes Measures-Principles of Good Practice” guidelines. The translation process included three steps: (1) two native English translators who were familiar with the examined topics have independently translated the questionnaire into Italian (forward translation); (2) one native Italian translator, aware of the subject matter, chose the best translation in order to create the final Italian version of the tool; and (3) a bilingual person with a certificated knowledge of the English language translated the text back into English (backward translation). The forward and backward translations were reviewed by qualified Italian and English-speaking clinicians before agreement on the final version. Final step consisted of having the approval of the Italian version by the main authors of the instrument.

Participants

Participants were recruited through a convenience sampling method, primarily via social media (e.g., WhatsApp groups, messaging function on social networks) and e-mails contacts, to facilitate broad and efficient data collection, from individuals across different regions of Italy. This procedure21,22 appeared appropriate, given the primary aim of the study. Participants were considered eligible based on the following inclusion criteria: (i) age over 18 years old, (ii) native Italian speaker or individual proficient in understanding the instructions and the questionnaire in Italian, (iii) access to an email address and internet, and (iv) self-declared absence of neurological or psychiatric disorders, confirmed through signed informed consent23. No gender-based restrictions were applied. Description of the sample structure and demographic characteristics is provided in the results section.

Procedures

The Italian version of the BIF questionnaire (Supplementary Information, S1) was administered using Google Forms. The link to the questionnaire was distributed via digital channels by project coordinators and collaborators, between the 19th September and the 20th October 2024. Responders were required to provide an email address, before starting to complete the BIF. This was necessary to send a follow-up link inviting participants to complete the questionnaire again, two weeks later, allowing for the assessment of test–retest reliability. Recruitment was conducted solely by the University of Genoa. The University of Udine and IRCCS E. Medea contributed to translation, data analysis, and manuscript writing.

Statistical analysis

Based on the literature, a sample of over 300 participants is considered good24 or adequate25 for validating measurement scales. Additionally, the COSMIN guidelines recommend a minimum of 7 participants per item-scale and at least 100 subjects for structural validity studies26. Finally, for assessing test–retest reliability and measurement error analysis, a sample size of 50 to 99 participants is considered satisfactory27.

Demographic characteristics of the sample were summarized using descriptive statistics, with mean and standard deviations for the continuous variables (age and education), and absolute frequencies with percentage for the dichotomous variables (gender).

Descriptive statistics for the total and for each BIF item were reported as frequency with percentage, whereas its psychometric properties were investigated specifically in the domains of validity and reliability as required by the COSMIN recommendation26.

Validity

The structural validity, defined as the degree to which an instrument’s scores accurately represent the dimensionality of the concept to be examined26, was analyzed to verify the one-dimensionality of the 25-items of the questionnaire through a Confirmatory Factorial Analysis (CFA) using the Weighted Least Squares Mean and Variance Adjusted (WLSMV) estimator, appropriate for the dichotomous survey items, within the Classical Test Theory (CTT). The analysis considered both absolute and relative indices related to a good fit model based on scaled statistics: chi-square test of model fit (χ2); Root Mean Square Error of Approximation (RMSEA) ≤ 0.06 for an excellent model fit, between 0.06 and 0.08 for an acceptable fit, with values up to 0.10 considered adequate28; Standardized Root Mean Square Residual (SRMR) deemed acceptable if ≤ 0.08; Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI), both considered excellent for values > 0.95, or acceptable for the range between 0.90 and 0.9529,30. In addition, standardized factor loadings (β) for each item were reported to assess the adequacy of each indicator in representing the latent construct. According to literature, a standardized loading of 0.40 or higher is considered acceptable, indicating that the item contributes substantially to the measurement of the underlying factor. To control for the influence of multivariate outliers, we computed the Mahalanobis Distance (MD) for each participant31,32 MD evaluate how far an individual’s response pattern deviates from the multivariate center of the sample distribution. A critical value of χ225, 0.001 = 52.619 was used, corresponding to 25 degrees of freedom (equal to the number of BIF items) at a significance level of p < 0.001. Observations exceeding this threshold were to be flagged as multivariate outliers.

Reliability

The reliability was assessed in terms of internal consistency to verify the homogeneity and the interrelatedness of the all 25 items of the scale. Both Cronbach’s alpha (α) and Donald’s omega (ω) coefficients, the last more accurate for binary items, were used as the reference index for the one-dimensionality. Both coefficients used to be considered acceptable with values between 0.70 and 0.79, good in the range between 0.80 and 0.89, and excellent with values > 0.9033,34,35. Values above 0.950 might indicate redundancy within items. Additionally, if an item was deleted, a lower value of the α or ω coefficient, compared to the total alpha, or to the total omega respectively, was expected36. In addition, at the item-level analysis, Inter-Item and Item-to-Total correlations (i.e., IIC and ITC) were assessed by the Spearman’s rho correlation coefficient according to the psychometric validation procedures for self-report instruments with dichotomous items37. The IIC assess if items are positively related in directionally expected ways, the ITC check if each item contributes to the overall construct measured by the scale.

Average IIC was considered acceptable for values ranging from 0.15 to 0.5025, while ITC was considered acceptable for ≥ 0.25, and excellent for coefficients > 0.3038.

Test–retest reliability was checked by administering the questionnaire twice, approximately 2 weeks apart. The Intraclass Correlation Coefficient (ICC) using a 2-way random effects model with a 95% confidence interval (CI) and an absolute effect was used to evaluate the stability of the BIF results as proper methods to account for both agreement and consistency between repeated measures. For group measurements, ICC values were classified as poor if below 0.50, moderate between 0.50 and 0.75, good between 0.75–0.90 and excellent above 0.9039,40.

Lastly, the degree of uncertainty associated with the measurement error of two assessments was further explored by calculating the Standardized Error of Measurement (SEM) and the Minimal Detectable Change (MDC)41. The SEM was calculated using the following formula: SD x\(\sqrt{\left(1-ICC\right)}\) where SD represents the standard deviation of the baseline data and ICC the value obtained from the test–retest reliability and the MDC \(\sqrt{\left(1-ICC\right)}\) which corresponds to the z-score equivalent to a 95% confidence level34,42.

BIF technical features, intended as feasibility and acceptability, were evaluated considering the ease of online dissemination and the average completion time by participants.

Data collected on the Forms platform were exported and organized in Excel files. Statistical analyses were performed using Jasp [v. 0.18.1]43 and Jamovi [v. 2.26]44 softwares.

Despite the brevity of the questionnaire and the voluntary nature of participation, we adopted both preventive and post-hoc strategies to mitigate the risk of careless responding, as recommended by Ward and Meade45. To reduce inattentive responding, we provided clear and standardized instructions in both the informed consent and on-screen prior to survey initiation. Completion of all items was mandatory to allow submission of the questionnaire. Additionally, we monitored completion times and excluded responses that were excessively fast or delayed, which could indicate insufficient engagement.

As part of post hoc quality control, we screened the dataset for response patterns indicative of low-effort behavior (e.g., consistently selecting the same binary option across all items). Furthermore, multivariate outlier analysis using Mahalanobis distance was conducted to identify atypical response profiles. These checks were performed prior to the psychometric analysis of the BIF.

Results

The process to define the pooled sample analyzed in this study is illustrated in Fig. 1. A total of 359 volunteers responded to the first administration of the questionnaire, and 232 responded to the second one. Participants had a mean age of 36.73 years (SD = 14.12), 35% (125 participants) of the sample were men, and the average level of education was 15.81 years (SD = 3.51).

Fig. 1
figure 1

Flow chart of the study. The figure summarizes the steps of the study: in the first stage of the validation process, a link to access to the questionnaire was sent by mail to participants (359 questionnaires obtained); in the last stage, a second link to the questionnaire was sent to analyze the test–retest reliability (232 questionnaires obtained).

Full size image

The mean total score, and the descriptive statistics for each item of the BIF are described in Table 1.

Table 1 Descriptive statistics of the Italian version of the Behavior Identification Form (BIF).
Full size table

Both genders showed a general tendency to interpret action with a high-level perspective, with 98 (78%) men and 154 (66%) women in particular choosing the more abstract option.

Structural validity analysis showed proper fit indices results to one-factor model from the CFA estimated with WLSMV. The one-dimensional model was expressed as χ2 = 334, df = 275; RMSEA = 0.025, 90% CI: 0.013—0.033; SRMR = 0.075; CFI = 0.977; and TLI = 0.975. The CFA revealed that item factor loadings ranged from 0.60 to 1.59 for unstandardized values and over 0.423 for standardized ones, with exception of items 19 (β = 0.39) and 22 (β = 0.28) that fall just below the cut-off. (Table 2). No multivariate outliers were identified based on the MD analysis, as no participant exceeded the critical value of χ2(25, 0.001) = 52.619. A remarkable good internal consistency (α = 0.857; ω = 0.858) was shown for the BIF (Table 2).

Table 2 Standardized (β) and unstandardized factor loadings (Estimate) for each item on the latent factor (F1) from the confirmatory factor analysis (CFA) estimated with WLSMV.
Full size table

The average of inter-item correlation was 0.192. From the analysis of the item-to-total correlations emerged coefficients higher than 0.25 for all items, except for item #22 (ITC = 0.197). Cronbach’s alpha-if-item-deleted was solid, as suggested that the deletion of almost all items did not increase α, ranged from 0.848 to 0.855, apart from item #22 (α = 0.858 or ω = 0.859 if item deleted). Similarly, values of Donald’s omega confirmed the result with other values between 0.849 to 0.856 (Table 3).

Table 3 Item descriptive statistics, and internal consistency with Cronbach’s α and Donald’s ω values of the Italian version of the BIF.
Full size table

Test–retest reliability was calculated on 232 subjects. Results showed an excellent reliability with ICC(2,1) = 0.925 95% CI = 0.903–0.942), and two measurement error indices with a SEM of 6.28 points and MDC of 6.95 points.

The Italian version of BIF was completed by the recruited participants in an average time of 5 min and 39 s. Only one missing questionnaire were found among the data extracted. No indications of careless or inattentive responding were found in the remaining data. No problems of comprehension were signaled, nor troubles with the digital platform.

Discussion

This study had the goal to validate the Italian version of the BIF and test its validity and reliability properties, representing the first investigation into the classic psychometric properties of BIF in Italian healthy adults. The Italian BIF revealed a good one-dimensionality fit for the structure validity, a good internal consistency, and an excellent test–retest reliability. From our study it could be infer that BIF could be used systematically to monitoring trends towards high or low levels of identification behaviors in Italian healthy adults.

Previous research reported the use of translated versions of the BIF in other languages, as Dutch46, Turkish18, Japanese 47, and Chinese48 but translated versions were not find into citing papers and no process of validation of the instrument in these languages was reported. Only a French translation, administered online to a student sample, was found. In this case participants preferred higher level responses and BIF revealed a good internal consistency value (α = 0.79)49. Furthermore, evidence reported an extensive validation process, with focus on validity and reliability properties, only in Polish50.

Collecting 359 corrected responses of participants at the first administration of BIF and 232 in the second phase, the achievement of desired sample size according to the recommendation of the scientific literature was obtained.

Structural validity data analysis confirmed a good unidimensional fit model regarding the Action Identification Theory, aligning with the expected theoretical constructs, as for the original and the Polish versions of the BIF. The value of RMSEA and SRMR resulted appropriated, below the recommended thresholds; in particular RMSEA showed a narrow confidence interval indicating precise estimation and excellent absolute fit. In addition TLI and CFI reached an excellent threshold. In the Polish study, the author also investigated the structural validity of the questionnaire through an Exploratory Factor Analysis that converged to a one factor structure. These considerations supported our decision to retain the original structure without post hoc re-specification. CFA results were not affected by multivariate outliers, as none were identified using MD analysis. This suggests that within our healthy adult sample, no response patterns significantly deviated from the multivariate distribution, supporting the robustness of the model estimation. Reliability analyses reported a good internal consistency with a total α value of 0.857, reinforced by the robust value of the 0.858-ω coefficient, confirming the unidimensional structure of the BIF. Value of α coefficient appeared in line with the ones reported in the original study (α = 0.84) and in Polish version (α = 0.85). The IIC (0.192) fell within the acceptable range of values indicating that items were well correlated and measuring the same construct without be redundant.

Despite the result of the item #22 below the level of acceptability, the ITC value of 0.41 (cut-off > 0.200) was good and greater than values (0.35) reported both by the original and the Polish study. The Cronbach’s α and the Donald’s ω, if an item was deleted, was lower than the overall value suggesting that all items contribute to the generation of the instrument-score. Both the α and the ω coefficients of item #22 (Travelling a car) were an exception, with values of 0.58 and 0.59 respectively, indicating an insignificant increment of 0.01, lower than critical increment of 0.1. A similar pattern with the unique increment of the Cronbach’s α in the ITC analysis was found in the Polish study50. Although Item19 (β = 0.393) and Item22 (β = 0.255) fall just below the conventional threshold of 0.40 for standardized factor loadings51,52, the overall model fit indices and internal consistency metrics (e.g., Cronbach’s alpha and McDonald’s omega) were within acceptable to good ranges. This suggests that, although some items are weaker indicators, the scale as a whole reliably measures the intended construct. These findings are consistent with previous research on complex psychological constructs52 and highlight the need to evaluate both item-level and scale-level performance when validating instruments.

Regarding the test–retest analysis, the Italian version of BIF revealed excellent reliability with a ICC(2,1) (0.925) slightly higher than that of the Vallacher’s study (0.91). In addition, we investigated two other parameters associated with reliability not explored in previous studies, such as measurement error, finding a relative low SEM (6.28) below the standard deviation value (22.92), and MDC (6.95) over an interval of about 2 weeks. This time period was set at the duration used in the original study, but some participants took up to 15 days longer to give second responses. However, we considered that the construct was not spontaneously alterable in that time frame. The strong psychometric properties demonstrated by the Italian version of BIF outlined the potential practical utility in both research and applied settings of this outcome instrument.

Finally, the limited resources required, in terms of cost and time, to complete BIF made the instrument feasible in a straightforward manner and in particular the possibility of the online administration suggests the possibility of remote and repeated assessments and new development in digital health research also in large-scale and with longitudinal design.

Limitations and further research

There are some limitations in this study that are worth discussing. First, participants were enrolled through a convenience sampling method, which is more exposed, than random one, to the risk of selection bias, because the sample might be characterized by some aspects that are not typical of the larger population. For example, the relatively low average age of the sample might have hidden some age-related effects, considering that changes in the sensorimotor system might exert some influences on the choice of action identification. In addition, the web-based dissemination of access to the questionnaire, might have influenced the selection of participants, considering that the elderly might still be unfamiliar with technology. We could have remedied these aspects enrolling a larger number of participants than our cohort, which, however, satisfied abundantly the recommendations of the COSMIN for psychometric analyses. Another issue is the small number of demographic variables collected, which did not allow for a better detailed characterization and description of the sample. For these reasons, future research should consider employing larger and more representative samples to confirm and to extend the psychometric properties of the BIF across diverse populations, adding more variables (e.g. income, area of residence) to facilitate subgroup analyses, enhance external validity and the robustness of conclusions. Finally, another limitation was the reliance on self-reported health status without formal clinical screening. Nevertheless, this approach aligns with common practice in large-scale psychometric validation research and represents a reasonable compromise between methodological standards and feasibility. Our study lacked criterion validity analysis, which was elaborated in the Polish study comparing other measures relevant to the personal agency construct. Considering that the assessment was administered online, we chose to capture only one outcome measure to avoid the risk of excessive drops out or missing data.

In addition to a more systematic investigation in elderly, future studies should consider validating the BIF questionnaire and analyzing its responsiveness in clinical settings. As suggested by Watkins15, a defective action identification mechanism in terms of high- or low- level description can be detected in a transdiagnostic way in a wide range of conditions including autism53. Even if, to the best of our knowledge, there are no studies testing this issue, Marsh and colleagues54 assessed the relationship between the score at the Autism Quotient55 and the score at BIF in a sample of healthy adults. Their findings showed an inverse relation between autistic traits and the proportion of high-level responses in the BIF, thus supporting the importance of investigating the action identification mechanisms also in clinical samples with autism. Furthermore, it should be investigated also patients with Parkinson’s Disease in which the discrepancies between different levels of action identification and personal agency seem to be relevant.

The impact of the construct of personal agency in individual assessment justifies the use of the BIF as a valid and reliable outcome measure in healthy Italian adults.

Conclusion

The Italian version of the BIF proved to be a valid and reliable tool applied on healthy adults to assess own level of abstractness in action identification. Our results were in line with the good psychometric properties of the original scale7 and with the Polish50 version. By adhering to COSMIN guidelines and employing comprehensive psychometric analysis, including CFA, internal consistency, test–retest reliability, and measurement error indices, we have enhanced the methodological rigor compared to earlier studies.

Further studies are required to evaluate the introduction of this instrument of measure in clinical practice (e.g., in Parkinson’s Disease and Autism Spectrum Disorders) to assess some relevant characteristics as embodiment and sense of agency.