ψ %HOJLFD 3V\FKRORJLFD Hernández, G. P. et al. (2020). A Brief Version of the Implicit Positive and Negative Affect Test (IPANAT-18). Psychologica Belgica, 60(1), pp. 315–327. DOI: https://doi.org/10.5334/pb.544 RESEARCH ARTICLE A Brief Version of the Implicit Positive and Negative Affect Test (IPANAT-18) G. P. Hernández*,†, S. Edo*, M. Quirin† and T. Rovira* As self-reports of affect are limited in several regards, an indirect measure of affect, the Implicit Positive and Negative Affect Test (IPANAT; Quirin, Kazén, & Kuhl, 2009) has previously been developed and adapted to more than 10 languages (Quirin et al., 2018), showing adequate reliability and validity. Based on a sample of 242 Spanish adults (111 males), we evaluate a trimmed 18 items version of the IPANAT (IPANAT-18). Item reductive procedures consisted in a random selection of the stimuli words used in the IPANAT. Psychometric properties of the IPANAT-18 were evaluated via Confirmatory Factor Analysis. In addition, correlational analyses were used to determine the relationship between the brief and the full version of the IPANAT, and with explicit measures of affect. We replicated a two-factors structure of positive affect versus negative affect and found a good fit for the IPANAT-18 model (CFI = 1; TLI = 1; RMSEA = .00; SRMR = .03). Reliability was adequate (implicit PA, α = .86; implicit NA, α = .77) and the pattern of relation- ships with explicit affect measures were congruent and consistent with previous findings. Differences between the mean scores of implicit affect assessed with 18 items or 36 items were statistically non-significant, and showed strong correla- tions (PA, r = .92, p < .01; NA, r = .88, p < .01). In sum, the IPANAT-18 showed satisfactory psychometric properties and constitutes a useful tool for economi- cally measuring affective processes such as in experimental and economical multi- ple assessment (e.g., daily diary) settings. Keywords: implicit affect; IPANAT; psychometric properties Traditionally, psychological assessment of it has been found that people do not always affective states usually relies on the individ- identify and report emotions accurately ual’s own report of their feelings. However, (Quirin, Kazen, & Kuhl, 2009). The latter may partly be attributed to the complexity of affective experiences, as they are comprised * Stress and Health Research Group (GIES). of different components such as situation Departament de Psicologia Bàsica, Evolutiva appraisal, subjective feelings, expressive i de l’Educació, Universitat Autònoma de Barcelona, Barcelona, ES behavior, physiological responses, and action † TUM School of Management, Technical Univer- preparation (Scherer & Moors, 2019). It has sity of Munich, Munich, DE been argued that these different processes Corresponding author: Gina Hernández occur at a pre-reflective (i.e., automatic) and (gina.hernandez.coronel@gmail.com) a reflective (i.e., rational) level (Lieberman, 316 Hernández et al: A Brief Version of the IPANAT 2019). Therefore, self-report methods may process expressed in the participants’ biased not fully reflect an individual’s affective expe- judgments. Accordingly, the IPANAT uses rience. Hence, the importance of studying participants’ ratings of the degree to which implicit (i.e., automatic) affective processes. six nonsense words (i.e., SAFME, VIKES, Implicit affective processes are in line with TUNBA, TALEP, BELNI, and SUKOV) sound a dual-process view of appraisal theories of like six mood adjectives (i.e., happy, cheer- affect (Clore & Ortony, 2000). According to ful, energetic, helpless, tense, and inhibited). this view, information can be processed with Thus, the test is composed of 36-items, which reflective propositions and rules (which are scored on a 4-point Likert scale, ranging convey one or more appraisal values) but from doesn’t fit at all to fits very well. alternatively (or additionally) be processed The IPANAT showed good psychometric in an associative way (automatically acti- properties and construct validation (Quirin vating learned associations between rep- et al., 2009; Quirin et al., 2018). In addi- resentations of the stimuli and previously tion, criterion-based validity was found by stored appraisal outputs) (Moors, 2013). research showing relationships between Accordingly, the affective experience would implicit NA and low implicit PA with slow initiate with a pre-reflective process with blood pressure recovery after harassment several simultaneous automatic processes (Brosschot et al., 2014; van der Ploeg et al., giving rise to experience that has not (yet) 2014), and under unconscious stress induc- been reflected on. In line with this approach tion (van der Ploeg et al., 2019). As well as of affect as information processing, implicit with both stress-contingent and circadian affect is conceptualized as the automatic saliva cortisol, which did not occur for explicit activation of cognitive representations of affect (Mossink et al., 2015; Quirin, Kazén, affective experiences (Quirin et al., 2009). Rohrmann, & Kuhl, 2009). An fMRI study Previous research has demonstrated that demonstrated that implicit (IPANAT) but not affective processes, even if not fully recog- explicit negative affect predicted accuracy of nized, can impact human behavior (e.g., recognizing briefly presented anger gestures, Winkielman et al., 2005), and are related to as well as concomitant neural correlates in brain processes (Lane, 2008; Pessoa, 2013), the fear network of the brain (Suslow et al., and health (e.g., Quirin & Bode, 2014; Lane, 2015; see also Quirin & Lane, 2012, for the 2008; Weil et al., 2019). A number of proce- necessity of considering implicit affect in the dures have been developed for taping affec- neurosciences). tive processes indirectly, such as the Implicit Bodenschatz et al. (2018) used eye-tracking Association Test (IAT; Greenwald et al., 2003; in a healthy population to demonstrate that see also IAT-Anxiety, Egloff, & Schmukle, implicit NA predicts attention towards sad 2002), the Affect Misattribution Procedure faces over and above self-reported depres- (AMP; Payne et al., 2005). However, these sive symptoms. Kazén et al. (2014) found measures have been developed to assess indi- that implicit NA predicted local processing, viduals’ attitudes (or self-concepts) rather whereas implicit PA predicted global process- than affect itself, which has been led to the ing in individuals with low versus high emo- development of the IPANAT. tion regulation abilities, respectively, these The IPANAT aim is to assess a pre-reflective effects were not found for explicit affect. (i.e., automatic) dimension of affect, and Additional studies demonstrated validity draws on the principle of affect infusion as of the IPANAT as an affect measure that is a method to assess implicit affect. According incremental to explicit affect (e.g., Dekker & to this principle, affect exerts an impact on Johnson, 2018; Quirin et al, 2011; Remmers evaluative processes influencing the judg- et al., 2016). Hence, implicit affect assessed ments of unrelated objects. Thus, the goal of via the IPANAT appears to contribute the the test is to capture the automatic affective understanding of affective phenomena. Hernández et al: A Brief Version of the IPANAT 317 In addition, the IPANAT has been adapted who provided consent were then given a to many languages, displaying good psy- URL directing them to the experiment. More chometric properties (e.g., Hernández et than 90% (i.e., 218) of participants reported al., 2020; Shimoda et al., 2014; Sulejmanov & to have been born in Spain. Regarding the Spasovski, 2017). Results from ten different education level, the majority of participants countries showed that the best-fitting model self-reported to have a university degree or consisted of two factors corresponding to above (52%). Otherwise, 37% reported a positive affect and negative affect (on aver- high school degree, and 11% reported a sec- age, χ2/df = 2.53, CFI = .96, TLI = .91). Both ondary school degree. factors showed a good reliability coefficient, on average, implicit PA, α = .81; implicit NA, Materials α = .78 (Quirin et al., 2018). Implicit Affect scale Investigations on affect and health often A Spanish version of the IPANAT was used require economical assessments. For exam- (see Hernández et al., 2020). All testing took ple, due to the fact that affective processes place online via Qualtrics (Qualtrics Provo, are fleeting after experimental affect induc- 2013). In total, the experiment took approxi- tion (see Hermans et al., 2001), because mately 10 minutes to complete. A comput- sometimes participants respond to the erized version of the IPANAT presented one IPANAT in multiple assessments (like in eco- item each per screen, after the presentation logical momentary assessment studies), or of the instruction (i.e., cover story) of the simply because it is administered in conjunc- IPANAT. Then, participants were asked to pro- tion with time consuming other measures. vide judgments of six artificial words across Therefore, the purpose of this study was to six mood adjectives. For each of the artifi- create and evaluate a brief version of the cial words (SAFME, TALEP, BELNI, SUKOV, original test (called IPANAT-18 in the remain- GOLIP, and KERUS) participants indicated der of this article). A validated brief version of on a four-point answer scale (1 = doesn’t fit the test could also improve the reliability on at all, 2 = fits somewhat, 3 = fits quite well, some experimental designs (e.g., if there is and 4 = fits very well) to what extend does the need of repeated measures of affect), as well sound of the artificial word convey each of as avoid extra burden or boredom to partici- the following moods: happy, helpless, ener- pants. Thus, a brief version would improve getic, tense, cheerful, and inhibited. Thus, the instrument’s utility without sacrificing the test consisted of 36-items. The artificial its psychometric properties. words were randomly presented to avoid order effects, each adjective within the same Method artificial word was also randomized, and the Participants six mood adjective belonging to each arti- The sample included 242 Spanish adults ficial word were presented subsequently. (111 males). Participants’ age after classi- Global scores for implicit PA and implicit fication into age bands of 18–24, 25–34, NA were computed by averaging the scores 35–44, 45–54, and 55–65 was distributed derived from positively valence, and nega- as follows: 18%, 18%, 26.8%, 18.9% and tively valence adjectives (following Quirin, et 18.3%. Participants were recruited online al., 2009). by a Spanish market research firm (CERES), they received 12 euros as compensation for Explicit affect scales their participation. The only requirement After answering the IPANAT participants for participation was to be above 18 years. were presented with a series of person- Participants first saw a full description of the ality and affect questionnaires used to experiment, which served simultaneously examine construct validity of the IPANAT. as the informed consent form. Participants Explicit PA and NA were assessed with two 318 Hernández et al: A Brief Version of the IPANAT instruments. First, we used the broadly via Cronbach’s alpha coefficient, to deter- applied Positive and Negative Affect Schedule mine the best ratio between the length of (PANAS, Watson et al., 1988; Spanish version: the test and good internal consistency. This Lopez et al., 2015). Second, explicit affect item reduction analysis based on classical was also assessed by asking participants for test theory was found to be a reliable item explicit mood judgments of the same mood reduction method (Erhart et al., 2010) in adjectives included in the IPANAT (i.e., ask- comparison with other methods like Rasch ing individuals to report the extent to which item-fit analysis. As suggested by Erhart et they feel happy, cheerful, energetic, helpless, al. (2010), our study accompanied this item tense, and inhibited at the moment) on a rat- reduction method by additional analysis ing scale from 0 (not at all) to 10 (absolutely) (i.e., confirmatory factor analysis) to cor- (following Quirin et al., 2009). Analogously roborate the psychometric properties of to the original IPANAT, we composed a PA the instrument. Once Cronbach’s alpha and an NA scale computing average scores coefficient provided a notion of the least for happy, cheerful, and energetic, versus number of items required to keep the psy- helpless, tense, and inhibited, respectively. chometric properties of the original IPANAT, item reductive procedure consisted of a ran- Statistical Analyses dom selection of the stimuli words used The goal of the present study was to create and in the IPANAT. Then, the newly stablished evaluate a brief version of the IPANAT. As other set of items were extracted for the original projective tests, the IPANAT uses judgments of 36-items. Finally, the descriptive statistics, artificial words to track changes on responses reliability coefficient, and latent structure to ambiguous stimuli with the objective of of the full IPANAT were compare with the revealing pre-reflective emotions. As detailed brief version. before, the instrument items are composed of As mentioned above, the latent struc- six mood adjectives that are assessed several ture of the IPANAT-18 was evaluated using times, then the 36 items are in fact six truly Confirmatory Factor Analysis (CFA). The CFA different items asked repeatedly to capture model tested was based on the model pro- biased responses. Thus, for the brief version posed by authors of the original test and pre- of the IPANAT it is paramount to identify the vious findings with the IPANAT (see Quirin number of repetitions of the items and not et al., 2018). The CFA model expressed the which particular items are needed to keep hypothesis that the IPANAT measures two in a brief version (since they are redundant), factors, implicit NA and implicit PA. Scores thus a random selection of the right number for each one of the six mood adjectives of items should yield similar psychometric assessed (i.e., 3 for PA and 3 for NA) were properties that the full test. As suggested by calculated by averaging across ratings of the Taber (2018), high levels of Cronbach’s alpha combination of the mood adjective and the indicate that items in a scale elicit the same three artificial words, then the correspond- pattern of responses (which implies they are ing 3 adjectives were loaded to its belong- redundant), even though a higher number of ing factor. It was a restricted model, which items in a scale improve the reliability, addi- allowed each of the items to load on the tional items measuring the same thing as the respective predicted factor only. Previous existing items leads to redundancy that is findings indicate that a correlation between inefficient, because almost no additional use- the two underlying factors could occur (see ful information is obtained, nonetheless the Quirin, et al., 2018). Thus, in our study the instrument takes longer to administer. two factors were set to be non-orthogonal, Since we aimed to improve the usefulness to better explore this possibility. According of the test, in our study reliability analysis to Izquierdo et al. (2014), to allow the covari- for different number of items were tested ance of the latent factors of the model is Hernández et al: A Brief Version of the IPANAT 319 the better way to corroborate its possible between the brief and the full version of the orthogonality. IPANAT, and with explicit measures of affect. The CFA models included error variances Basic statistical analyses were conducted for each item and were set to load with a using IBM SPSS Statistics 22.0. In addition, coefficient of 1. We estimated factor load- Confirmatory Factor Analysis (CFA) were per- ings via diagonally weighted least squares formed using R 3.6 and RStudio 1.2. (DWLS) estimator, which has specifically been designed for ordinal data (Cheng- Results Hsien, 2016). We used Chi-squared values An analysis of the internal consistency of and degrees of freedom for each model to the IPANAT’s scales while performing item’s assess the fit of the CFA models. As well as reduction (see Figure 1), determined that Comparative Fit Index (CFI; Bentler, 1990), the best ratio between number of items the TLI (Tucker-Lewis index), the Root-Mean- and an acceptable level of alpha coefficient Square Error of Approximation (RMSEA), and were three artificial words (i.e., 18 items). Standardized Root Mean Square Residual Since the alpha coefficients for 18 items (SRMR), as they are commonly recommended corresponded to the least number of items to assess absolute measures of fit (Browne & with similar reliability coefficient to the ones Cudeck, 1992; Jackson et al., 2009; Steiger & reported for the different versions of the full Lind, 1980). Following guidelines for Hopper IPANAT (see Quirin et al., 2018). Therefore, et al. (2008), the present study used the next for the IPANAT-18 three artificial words were thresholds for determining model fit: Chi- randomly selected (i.e., SAFME, TALEP and squared (CMIN/df) less than 3, CFI ≥ 0.95, BELNI) from the stimuli words used in the TLI ≥ 0.95, RMSEA ≤ 0.05 and SRMR ≤ 0.08. IPANAT-S. Finally, we used correlational analysis After having completed the test, partici- and Z-tests to determine the relationships pants responded to a question about the Figure 1: Reliability of the IPANAT’s scales showed by items reduction. 320 Hernández et al: A Brief Version of the IPANAT presumed underlying aim of the IPANAT. 230, we have an acceptable ratio of 17.69 Twelve individuals suggested that the participants to 1 parameter estimated. test might assess affective states and were As Table 1 shows, the mean scores for PA excluded from the initial sample of 242 par- are higher than the mean score for implicit ticipants (4.95% of the sample). Descriptive NA. The latter is consistent with previ- statistics (mean scores, standard deviations, ous findings with the IPANAT (Quirin et al., skewness, and kurtosis) for the brief and the 2018). Table 1 also shows that the internal full version of the IPANAT can be found in consistency estimates for the IPANAT-18 Table 1. There were no missing data. After scales reached an acceptable level, implicit evaluating the assumptions of multivariate PA obtained an alpha coefficient of .86, while normality and linearity, we identified that implicit NA was .77. Moreover, the alpha coef- the assumption of multivariate normality is ficients are comparable to the ones reported slightly violated in our sample. Therefore, we by the original version of the test (Quirin et used the diagonally weighted least squares al., 2009). (DWLS) estimator, since this method pro- vides more accurate parameter estimates Factor Analysis (Mîndrilă, 2010). Regarding sample size, it The model tested for the brief version of the was determined that the size we used in the IPANAT-18 obtained a χ2 of 3.93, 8 degrees present study is adequate for the stability of freedom, a χ2/df (CMIN) of 0.49, with a of the parameter estimates, since 10 par- CFI of 1, the TLI was also 1, the RMSEA was ticipants per estimated parameter appears 0.00, while the SRMR was 0.02. According to be the general consensus (see Schreiber to Hu and Bentler (1999), those values indi- et al., 2006). In the CFA model we specify cate a good fit between the model and the 6 regressions, 1 covariance, and 6 variances, observed data (see also Schreiber et al., 2006). totalling 13 parameters that need to be esti- Table 2 depicts the χ2 and fit indices of the mated. Since we have a final sample size of full and brief version of the test, and Table 3 depicts standardized and unstandardized Table 1: Descriptive statistics and reliability coefficient of the brief and full version of the Implicit Positive and Negative Affect Test. M SD SK K α Implicit PA (Full version) 1.82 0.58 0.20 –0.89 0.91 Implicit NA (Full version) 1.59 0.44 0.60 –0.41 0.87 Implicit PA (IPANAT-18) 1.82 0.61 0.19 –1.00 0.86 Implicit NA (IPANAT-18) 1.57 0.46 0.78 0.42 0.77 Note: n = 230. Table 2: Fit Indices of Models Tested in Confirmatory Factor Analysis (n = 230). Model χ2 (df) χ2 /df CFI TLI RMSEA SRMR 1. IPANAT 3.23(8) 0.40 1 1 0.00 0.02 2. IPANAT-18 3.93(8) 0.49 1 1 0.00 0.03 Note: 1 = Full-IPANAT (36 items), restricted bi-factorial model (Implicit Positive/Negative affect), not allowing for cross loadings between factors; 2 = IPANAT-18 (18 items), same structure than model 1; CFI = comparative fit index; TLI = Tucker-Lewis index; RMSEA = root mean square error of approxi- mation; SRMR = standardized root mean square residual. Hernández et al: A Brief Version of the IPANAT 321 Table 3: Standardized and Unstandardized Coefficients for CFA Model 1(IPANAT) and Model 2(IPANAT-18) (n = 230). Observed variable Latent IPANAT IPANAT-18 construct β B SE β B SE Happy (Feliz) PA 0.94 1.00 0.88 1.00 Energetic (Activo) PA 0.92 1.03 0.07 0.87 0.94 0.08 Cheerful (Alegre) PA 0.89 0.97 0.07 0.90 0.99 0.09 Helpless (Desamparado) NA 0.75 1.00 0.68 1.00 Tense (Tenso) NA 0.83 1.23 0.11 0.79 1.24 0.16 Inhibited (Inhibido) NA 0.85 1.26 0.11 0.77 1.22 0.16 Figure 2: Results from Confirmatory Factor Analysis (model 2) for IPANAT-18 (n = 230). coefficients of the CFA Models. Along with full version was t(229) = –.35, p > .05, and Figure 2, the results suggest an acceptable for implicit NA t(229) = 1.22, p > .05. In addi- model fit for a two-factorial solution of the tion, implicit affect mean scores assessed IPANAT-18. Moreover, the fit indices obtained with the 18-items and 36-items versions by the brief version (18-items) are slightly showed strong correlations (implicit PA, lower, yet comparable to fit indices found r = .92; implicit NA, r = .88). for the full version on this sample, and to the ones reported for ten different versions of Relationships between IPANAT-18 and the full test (see Quirin et al., 2018). explicit scales of affect As shown in Table 4, the correlations Relationships between scales of the brief between the IPANAT-18 and explicit affect and full versions of the IPANAT measures are of moderate strength. In The differences between the mean scores of addition, Z-tests were run to compare the implicit affect assessed with 18 items or 36 correlations between implicit and explicit items were statistically non-significant. For scales of affect. For implicit negative affect, example, differences for implicit PA brief and the results show that the correlation 322 Hernández et al: A Brief Version of the IPANAT Table 4: Pearson correlations among Implicit Affect (IPANAT-18), Explicit affect (PANAS), and Explicit scale (Same Adjectives than on IPANAT). Measure IPANAT-18 IPANAT-18 Implicit PA Implicit NA PANAS PA 0.15* 0.07 ns Explicit scale PA (Same Adjectives) 0.26** 0.08 ns PANAS NA 0.15* 0.29** Explicit scale NA (Same Adjectives) –0.05 ns 0.15*** Note: n = 230, ** p < .05, *** p < .01, ns = non-significant. with explicit negative affect (assessed by results section, chi-square and fit indexes PANAS) is significantly higher than the corre- indicated a good fit of the proposed model. lation with explicit positive affect, z = 2.441, In addition, the sample size used in the pre- p < .01. Inversely, it was found that implicit sent study was adequate to produce relative positive affect (assessed by Same Adjectives stability of the parameter estimates. Internal Scale), was more strongly correlated to consistency analyses showed a good reliabil- explicit positive affect than to explicit nega- ity for both scales, and the CFA goodness of tive affect, z = 3.107, p < .01. fit was comparable to findings from previ- ous validations of explicit affect instruments Different versions of the IPANAT-18 (López et al., 2015). Statistical analysis were also performed on In our study, PA and NA dimensions the non-selected stimuli words of the IPANAT occurred to be non-orthogonal, as also (i.e., SUKOV, GOLIP and KERUS). Results indi- reflected in a positive correlation between cated that this set of three artificial words mean values of implicit PA and implicit NA. also shows good psychometric properties. This is consistent with previous findings from The fit indices of the CFA model of this brief cross-cultural studies with the IPANAT (see version were CMIN 0.52, CFI 0.99, TLI 0.99, Hernández et al., 2020; Quirin et al., 2018). RMSEA 0.01, and SRMR 0.04. The authors argued that positive correla- tions between positive and negative affect Discussion could be due the fact that different cultures The present study aimed to create and vali- attribute slightly different meaning to mood date a brief version of the IPANAT, a meas- adjectives. The latter is also consistent with ure for the indirect assessment of affect. findings of adjectives referring to personal- Based on the results from the items reduc- ity (Nye et al., 2008). In concordance, previ- tion procedure, three artificial words (i.e., ous cross-cultural studies with the IPANAT SAFME, TALEP and BELNI) were randomly showed that high correlations between selected from the six stimuli words used positive and negative affect was mostly due in the IPANAT. Therefore, the brief version to a positive correlation between the mood of the IPANAT is composed of 18 items. We adjectives energetic and tense (Quirin et al., explored the goodness of fit of IPANAT-18 via 2018). In addition, it has been argued that in CFA technique and found that the best fit- some languages the mood adjectives provide ting model supports a two-factor structure a smaller variability on the responses range. of the test, corresponding to implicit PA and Therefore, future studies exploring this implicit NA, which is in line with the factor hypothesis should use a sample with a strong structure found in the original IPANAT (see emotional context or under emotional prim- Quirin et al., 2009). As mentioned in the ing. Nonetheless, according to Brown (2006) Hernández et al: A Brief Version of the IPANAT 323 a factor structure with a positive correlation could reduce anchoring effects on partici- between factors might be the better model pant’s responses. fit, particularly if the factor loadings are In conclusion, the present study suggests strong, and the fit indices are better that the that the psychometric properties of the one-factor model, as previously found in the IPANAT-18 version are almost as good as IPANAT’s CFAs (see Hernández et al., 2020). those of the full-length measure. Hence, it Not least, convergent and discriminant appears that the shorter measure will serve validity of the IPANAT-18 was supported by studies requiring less time for administra- valence-congruent findings of correlations tion than the original test. The latter is espe- with explicit affect scales. For example, cially important for research where affective results showed that correlations between processes are experimentally induced, since the IPANAT-18 and explicit affect measures it has been determined that the induced are significant and of moderate strength. affect is often fleeting (Hermans et al., 2001), These moderate correlations are consistent so a brief version is useful to better capture with results reported for the original IPANAT, these processes. Likewise, research using since Quirin et al. (2009) reported significant repeated assessment, as daily-diaries studies, correlations of .20 for implicit and explicit can also benefit by an economical multiple PA and .22 for implicit and explicit NA. The assessment, since a shorter version of instru- moderate correlations between implicit and ments will help not to frustrate participants. explicit measures are also consistent with findings of other implicit measures like: the Acknowledgements Implicit Association Test (Greenwald et al., This work was partially made possible 2003), or the Affect Misattribution Procedure through a grant from the Templeton Rlg. (Payne et al., 2005) (See Echebarria-Echabe, Trust (TRT 0119) supporting MQ and GPH; 2013). According to some authors, these low by the National Council for Science and correlations between implicit and explicit Technology of Mexico (CONACyT) supporting measures can be due to different aspects, GPH, and the Spanish Government (under as motivational biases in the explicit meas- Grant PSI2016-76411-R) supporting GPH, SE ure, lack of introspective access of the par- and TR. Special thanks to Cafer Bakac for pro- ticipants, or even complete independence of viding advice regarding analysis. the underlying constructs (Hofmann et al., 2005). In addition, evidence of discriminant Competing Interests validity of the IPANAT-18 can be obtain for The authors have no competing interests to our results, since Z-tests showed that implicit declare. NA was more strongly correlated with explicit NA measures than with explicit PA measures, References the opposite was found for implicit PA. Bentler, P. M. (1990). Comparative fit indexes Finally, a different set of the artificial words in structural models. Psychological Bulle- (i.e., SUKOV, GOLIP and KERUS) can be used tin, 107(2), 238–246. DOI: https://doi. as a different version of the IPANAT-18. Since org/10.1037/0033-2909.107.2.238 results showed that the random selection of Bodenschatz, C. M., Skopinceva, M., items (i.e., three artificial words by 6 mood Kersting, A., Quirin, M., & Suslow, ­ adjectives) yield similar psychometric prop- T. (2018). Implicit negative affect pre- erties than the full test. The latter is useful dicts attention to sad faces beyond self- for researchers of the affective phenomena, reported depressive symptoms in healthy particularly in experimental settings were individuals: An eye-tracking study. Psychi- repeated measures of the test are needed, atry Research, 265, 48–54. DOI: https:// since having different version of the test doi.org/10.1016/j.psychres.2018.04.007 324 Hernández et al: A Brief Version of the IPANAT Brosschot, J. F., Geurts, S. A. E., Kruizinga, U., & European KIDSCREEN Group. I., Radstaak, M., Verkuil, B., Quirin, M., (2010). A comparison of Rasch item‐fit & Kompier, M. A. J. (2014). Does Uncon- and Cronbach’s alpha item reduction scious Stress Play a Role in Prolonged Car- analysis for the development of a Qual- diovascular Stress Recovery? Stress and ity of Life scale for children and adoles- Health, 30(3), 179–187. DOI: https://doi. cents. Child: care, health and develop- org/10.1002/smi.2590 ment, 36(4), 473–484. DOI: https://doi. Brown, T. A. (2006). Confirmatory factor org/10.1111/j.1365-2214.2009.00998.x analysis for applied research. New York: Greenwald, A., Nosek, B., and Banaji, M. Guilford Publications. DOI: https://doi. (2003). Understanding and using the org/10.5860/CHOICE.44-2769 implicit association test: I. An improved Browne, M. W., & Cudeck, R. (1992). scoring algorithm. Journal of Personal- Alternative Ways of Assessing Model ity and Social Psychology, 85, 197–216. Fit. Sociological Methods & Research, DOI: https://doi.org/10.1037/0022- 21(2), 230–258. DOI: https://doi. 3514.85.2.197 org/10.1177/0049124192021002005 Hermans, D., De Houwer, J., & Eelen, P. Cheng-Hsien, L. (2016). Confirmatory fac- (2001). A time course analysis of the tor analysis with ordinal data: Comparing affective priming effect. Cognition & Emo- robust maximum likelihood and diago- tion, 15(2), 143–165. DOI: https://doi. nally weighted least squares. Behavior org/10.1080/02699930125768 research methods, 48(3), 936–949. DOI: Hernández, G. P., Rovira, T., Quirin, M., https://doi.org/10.3758/s13428-015- & Edo, S. (2020). A Spanish Adaptation 0619-7 of the Implicit Positive and Negative Clore, G. L., & Ortony, A. (2000). Cognitive Affect Test (IPANAT). Psicothema, 32(2). Neuroscience of Emotion. In Series in DOI: https://doi.org/10.7334/psico- Affective Science. Cognitive Neuroscience thema2019.297 of Emotion, R. D. R. Lane, L. Nadel, G. L. Hofmann, W., Gawronski, B., Gschwend- Ahern, J. Allen, & A. W. Kaszniak (Eds.). (pp. ner, T., Le, H., & Schmitt, M. (2005). 24–61). Oxford: Oxford University Press. A Meta-Analysis on the Correlation Dekker, M. R., & Johnson, S. L. (2018). Major Between the Implicit Association Test depressive disorder and emotion-related and Explicit Self-Report Measures. Per- impulsivity: Are both related to cogni- sonality and Social Psychology Bulletin, tive inhibition? Cognitive Therapy and 31(10), 1369–1385. DOI: https://doi. Research, 42(4), 398–407. DOI: https:// org/10.1177/0146167205275613 doi.org/10.1007/s10608-017-9885-2 Hooper, D., Coughlan, J., & Mullen, M. R. Echebarria-Echabe, A. (2013). Relationship (2008). Structural equation modelling: between implicit and explicit measures guidelines for determining model fit. The of attitudes: The impact of application Electronic Journal of Business Research conditions. Europe’s Journal of Psychol- Methods, 6, 53–60. ogy, 9(2), 231–245. DOI: https://doi. Hu, L., & Bentler, P. M. (1999). Cutoff org/10.5964/ejop.v9i2.544 criteria for fit indexes in covariance Egloff, B., & Schmukle, S. (2002). Predictive structure analysis: Conventional crite- Validity of an Implicit Association Test for ria versus new alternatives. Structural Assessing Anxiety. Journal of Personality Equation Modeling: A Multidisciplinary and Social Psychology, 83(6), 1441–1455. Journal, 6(1), 1–55. DOI: https://doi. DOI: https://doi.org/10.1037//0022- org/10.1080/10705519909540118 3514.83.6.1441 Izquierdo, I., Olea, J., & Abad, F. J. (2014). Erhart, M., Hagquist, C., Auquier, P., Exploratory factor analysis in validation Rajmil, L., Power, M., Ravens‐Sieberer, studies: Uses and recommendations. Psic- Hernández et al: A Brief Version of the IPANAT 325 othema, 26(3), 395–400. DOI: https:// Nye, C. D., Roberts, B. W., Saucier, G. & doi.org/10.7334/ psicothema2013.349 Zhou, X. (2008). Testing the meas- Jackson, D. L., Gillaspy, J. A., & Purc-­ urement equivalence of personality Stephenson, R. (2009). Reporting prac- adjective items across cultures. Journal tices in confirmatory factor analysis: An of Research in Personality, 42, 1524– overview and some recommendations. 1536. DOI: https://doi.org/10.1016/j. Psychological Methods, 14, 6–23. DOI: jrp.2008.07.004 https://doi.org/10.1037/a0014694 Payne, B. K., Cheng, C. M., Govorun, O., Kazén, M., Kuhl, J., & Quirin, M. (2014). & Stewart, B. D. (2005). An inkblot Personality Interacts with Implicit Affect for attitudes: affect misattribution as to Predict Performance in Analytic vs. implicit measurement. Journal of Person- Holistic Processing. Journal of Personal- ality and Social Psychology, 89, 277–293. ity, 83(3), 251–261. DOI: https://doi. DOI: https://doi.org/10.1037/0022- org/10.1111/jopy.12100 3514.89.3.277 Lane, R. D. (2008). Neural substrates of Pessoa, L. (2013). The Cognitive-Emo- implicit and explicit emotional processes: tional Brain. From Interactions to a unifying framework for psychosomatic Integration. Cambridge: MIT Press. medicine. Psychosomatic Medicine, 70(2), DOI: https://doi.org/10.7551/mit- 214–231. DOI: https://doi.org/10.1097/ press/9780262019569.001.0001 PSY.0b013e3181647e44 Qualtrics Research Suite. (2013). Qualtrics Lieberman, M. D. (2019). Boo! The con- and all other Qualtrics product or ser- sciousness problem in emotion. Cog- vice names are registered trademarks or nition and Emotion, 33, 24–30. DOI: trademarks of Qualtrics, Provo, UT, USA. https://doi.org/10.1080/02699931.201 Retrieved from http://www.qualtrics. 8.1515726 com. López-Gomez, I., Hervas, G., & Vázquez, Quirin, M., & Bode, R. C. (2014). An Alter- C. (2015). An adaptation of the positive native to Self-Reports of Trait and State and negative affect schedules (PANAS) Affect The Implicit Positive and Negative in a Spanish general sample. Behavioral Affect Test (IPANAT). European Journal of Psychology-Psicología Conductual, 23(3), Psychological Assessment, 30(3), 231–237. 529–548. DOI: https://doi.org/10.1027/1015- Mîndrila, D. (2010). Maximum likelihood 5759/a000190 (ML) and diagonally weighted least Quirin, M., Bode, R. C., & Kuhl, J. (2011). squares (DWLS) estimation procedures: Recovering from negative events by A comparison of estimation bias with boosting implicit positive affect. Cogni- ordinal and multivariate non-normal tion and Emotion, 25(3), 559–570. DOI: data. International Journal of Digital https://doi.org/10.1080/02699931.201 Society, 1(1), 60–66. DOI: https://doi. 0.536418 org/10.20533/ijds.2040.2570.2010.0010 Quirin, M., Kazén, M., & Kuhl, J. (2009). Moors, A. (2013). On the causal role When nonsense sounds happy or help- of appraisal in emotion. Emotion less: The Implicit Positive and Negative Review, 5, 132–140. DOI: https://doi. Affect Test (IPANAT). Journal of Personal- org/10.1177/1754073912463601 ity and Social Psychology, 97(3), 500–516. Mossink, J. C. L., Verkuil, B., Burger, A. DOI: https://doi.org/10.1037/a0016063 M., Tollenaar, M. S., & Brosschot, J. Quirin, M., Kazén, M., Rohrmann, S., & F. (2015). Ambulatory assessed implicit Kuhl, J. (2009). Implicit but not explicit affect is associated with salivary cortisol. affectivity predicts circadian and reactive Frontiers in Psychology, 6. DOI: https:// cortisol: Using the implicit positive and doi.org/10.3389/fpsyg.2015.00111 negative affect test. Journal of Personality, 326 Hernández et al: A Brief Version of the IPANAT 77(2), 401–426. DOI: https://doi. tive affect test (IPANAT-M). Review of Psy- org/10.1111/j.1467-6494.2008.00552.x chology, 24(1–2), 39–44. Quirin, M., & Lane, R. D. (2012). The construc- Suslow, T., Ihme, K., Quirin, M., Lichev, V., tion of emotional experience requires the Rosenberg, N., Bauer, J., … Lobsien, D. integration of implicit and explicit emo- (2015). Implicit affectivity and rapid pro- tional processes. Behavioral and Brain Sci- cessing of affective body language: An ences, 35(3), 159–160. DOI: https://doi. fMRI study. Scandinavian Journal of Psy- org/10.1017/S0140525X11001737 chology, 56(5), 545–552. DOI: https:// Quirin, M., Wróbel, M., Norcini Pala, A., doi.org/10.1111/sjop.12227 Stieger, S., Brosschot, J., Kazén, M., … Taber, K. S. (2018). The use of Cronbach’s Kuhl, J. (2018). A cross-cultural validation alpha when developing and reporting of the implicit positive and negative affect research instruments in science edu- test (IPANAT). European Journal of Psycho- cation. Research in Science Education, logical Assessment, 1–12. DOI: https://doi. 48(6), 1273–1296. DOI: https://doi. org/10.1027/1015-5759/a000315 org/10.1007/s11165-016-9602-2 Remmers, C., Topolinski, S., & Koole, S. van der Ploeg, M. M., Brosschot, J. F., L. (2016). Why being mindful may have ­Quirin, M., Lane, R. D., & Verkuil, B. more benefits than you realize: Mindful- (2019). Inducing unconscious stress: Sub- ness improves both explicit and implicit liminal anger and relax primes show sim- mood regulation. Mindfulness, 7(4), ilar cardiovascular activity patterns. Jour- 829–837. DOI: https://doi.org/10.1007/ nal of Psychophysiology. DOI: https://doi. s12671-016-0520-1 org/10.1027/0269-8803/a000247 Scherer, K. R., and Moors, A. (2019). The van der Ploeg, M. M., Brosschot, J. F., & emotion process: event appraisal and Verkuil, B. (2014). Measuring Uncon- component differentiation. Annual scious Stress: the Implicit Positive and Review of Psychology, 70, 719–745. Negative Affect Test and Cardiovascular DOI: https://doi.org/10.1146/annurev- Activity After Anger Harassment. Psycho- psych-122216-011854 somatic Medicine, 76(3), A90–A91. Schreiber, J. B., Nora, A., Stage, F. K., Bar- Watson, D., Clark, L. A., & Tellegen, A. (1988). low, E. A., & King, J. (2006). Report- Development and validation of brief meas- ing structural equation modeling and ures of positive and negative affect: The confirmatory factor analysis results: PANAS scales. Journal of Personality and A review. The Journal of educational Social Psychology, 54(6), 1063–1070. DOI: research, 99(6), 323–338. DOI: https:// https://doi.org/10.1111/sjop.12227 doi.org/10.3200/JOER.99.6.323-338 Weil, A-S., Hernández G. P., Suslow, T., & Shimoda, S., Okubo, N., Kobayashi, M., Quirin, M. (2019). Implicit Affect and Sato, S., & Kitamura, H. (2014). An Autonomous Nervous System Reactions: attempt to construct a japanese version A Review of Research Using the Implicit of the implicit positive and negative Positive and Negative Affect Test. Fron- affect test (IPANAT). Japanese Journal of tiers in Psychology, 10, 1634. DOI: https:// Psychology, 85(3), 294–303. DOI: https:// doi.org/10.3389/fpsyg.2019.01634 doi.org/10.4992/jjpsy.85.13212 Winkielman, P., Berridge, K. C., & Steiger, & Lind, A. (1980). Statistically based Wilbarger, J. L. (2005). Unconscious tests for the number of common factors. affective reactions to masked happy In the annual meeting of the Psychometric versus angry faces influence consump- Society. Iowa City, IA. tion behavior and judgments of value. Sulejmanov, F., & Spasovski, O. (2017). Psy- Personality and Social Psychology Bul- chometric properties of the macedonian letin, 31(1), 121–135. DOI: https://doi. version of the implicit positive and nega- org/10.1177/0146167204271309 Hernández et al: A Brief Version of the IPANAT 327 How to cite this article: Hernández, G. P., Edo, S., Quirin, M., & Rovira, T. (2020). A Brief Version of the Implicit Positive and Negative Affect Test (IPANAT-18). Psychologica Belgica, 60(1), 315–327. DOI: https://doi.org/10.5334/pb.544 Submitted: 05 March 2020 Accepted: 17 August 2020 Published: 16 September 2020 Copyright: © 2020 The Author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/4.0/. Psychologica Belgica is a peer-reviewed open access journal OPEN ACCESS published by Ubiquity Press.