RETRIEVAL ABILITY AND DIVERGENT THINKING

Divergent thinking can be conceptualized as a cognitive ability that stresses creative fluency (Runco, 2020) as well as the originality of responses (Carroll, 1993). Other aspects have been proposed, such as the flexibility (Weiss & Wilhelm, 2022) and the elaboration of generated ideas (Forthmann, Oyebade, Ojo, Günther, & Holling, 2018), but are only rarely studied (Hornberg & Reiter-Palmon, 2017; Reiter-Palmon, Forthmann, & Barbot, 2019).

Creative fluency is theoretically intertwined with the ability to retrieve information. This ability has been defined as the “rate and fluency at which individuals can produce and selectively and strategically retrieve verbal and nonverbal information or ideas stored in long-term memory” (pp 102, Schneider & McGrew, 2018). The originality of ideas, however, relies on retrieval since it is based upon further processing of both stored and retrieved information, but retrieving information/ideas is not the first, foremost aspect. Previously, the retrieval part has been rarely considered a crucial aspect of creative cognition, because participants are deemed equally able to rely on information available in long term memory. However, novel methodological approaches in modeling semantic memory and accordingly retrieval processes, for example by semantic networks, have revived the link between goal-directed retrieval and retrieval that is rather based on free associations from memory as it is crucial in thinking divergently (Beaty & Kenett, 2023). Furthermore, research provides considerable evidence of how having a lot of ideas (creative fluency/retrieval ability) is related to finding the best idea. For instance, chance models of creativity, like the Equal Odds baseline, assume that creative hits are a positive linear function of the number of produced ideas (Forthmann, Szardenings, & Dumas, 2021).

Models of individual differences in human intelligence locate retrieval, creative fluency, and originality as aspects of idea production which is in term deemed part of the ability to retrieve information from long-term memory (e.g., Goecke, Weiss, & Wilhelm, 2024; Silvia, Beaty, & Nusbaum, 2013). More specifically, Schneider and McGrew (2018) propose five first-order factors in the area of idea production: ideational fluency (frequency of coming up with ideas, words, phrases), expressional fluency (expressing ideas in various ways), associational fluency (producing quality ideas to a given concept), alternative solution fluency (coming up with alternative solutions to a given problem) and originality/creativity (producing original ideas regarding a specific topic).

In sum, embedding divergent thinking in models of intelligence—most prominently below a factor of retrieval ability (Schneider & McGrew, 2018)—has a long tradition, and the significant role of retrieval of information from long-term memory in divergent thinking seems self-evident (Carroll, 1993; Guilford, 1956; Miroshnik, Forthmann, Karwowski, & Benedek, 2023). Figure 1 schematically displays the relationship between the three constructs and highlights methodological kinship in terms of performance appraisal.

As displayed in Figure 1, retrieval ability and creative fluency, can both be evaluated by using count variables. This implies that retrieval ability and creative fluency are predominantly assessed by human counts of correct and appropriate responses. Furthermore, as depicted in Figure 1 tasks that are instructed for providing the most original answer can be used to assess the construct of originality. However, in the literature these originality ratings are often also applied on creative fluency tasks that were initially instructed in terms of fluency (“Names as many (creative)”; see Scoring in Figure 1). This might be even the predominant application of tasks in creativity literature (e.g., creative fluency tasks that are instructed for creative fluency, but scored with regards to the count of correct answers and the originality), while tasks that only ask to name the most creative thing are less frequently applied (e.g., Saretzki et al., 2024). Scoring a creative fluency task for both, count data and an originality rating might be a time-efficient variant, that however is also to be considered critically given the dependency between the two scores that are derived from the same set of answers. The possibility to capture originality—as indicated by uncommonness, remoteness, and cleverness of ideas (Wilson, Guilford, & Christensen, 1953)—and not only idea production makes divergent thinking tasks valuable for creativity assessment (Runco, 2020; Weiss et al., 2021b). Although operationalizing creativity by means of originality seems to meet the core definition of creativity best, the literature on creativity is saturated with simple fluency assessments (count of appropriate answers) in divergent thinking tasks, as they are more time-efficient and less costly to score than human codings of originality (Forthmann, Goecke, & Beaty, 2023; Hornberg & Reiter-Palmon, 2017). Therefore, there are two important research questions that cannot be answered conclusively so far. First, are creative fluency assessments based on count variables psychometrically distinguishable from retrieval ability that is also based on count variables of appropriate responses? Depending on the answer to this question, a second important research question is, whether or not we find differential relations with established cognitive abilities with retrieval and creative fluency.

Despite the relative importance of retrieval ability for divergent thinking and the overt overlaps in scoring procedures (c.f., count of appropriate answers), only few studies explicitly addressed the relationship between divergent thinking and retrieval ability. In previous meta-analytic evidence, the simplex displayed in Figure 1 has even mostly been neglected (Acar, Runco, & Park, 2020; Said-Metwaly, Fernández-Castilla, Kyndt, & Van Den Noortgate, 2020). Such meta-analytic studies have, for example, focused on the impact of instruction and time constraints on the ability to be creatively fluent and original, but they have not taken the ability to retrieve information into account.

In contrast, other studies reported the relationship between aspects of divergent thinking and retrieval ability. For example, one multivariate study showed that retrieval ability (including a higher-order model with six narrow ability factors) predicted a factor capturing both creative fluency and originality significantly (β _originality = .44, β _fluency = .40; c.f., table 4 in Silvia et al., 2013). A similar correlation between divergent thinking and retrieval (r = .54), also on the latent level, was reported by Forthmann et al. (2019). Further evidence for this relationship was reported in a recent meta-analysis (r = .48, 95% CI: [.40, .56]; Miroshnik et al., 2023). This meta-analysis included 41 studies that were—among others—also coded for the type of instruction and the type of divergent thinking indicator that was used. Both variables were examined as moderators: the type of instruction did not play a significant role in the relationship between divergent thinking and retrieval ability, only the type of divergent thinking indicator played a significant role as a moderator in this relationship (Miroshnik et al., 2023). Another moderator, called Gr-ness (i.e., time restriction, instruction condition, and scoring condition of divergent thinking tasks), which should depict the conceptual overlap between Gr and DT measures, did not alter the relationship between divergent thinking and retrieval ability. Based on these results, the authors concluded that the communalities between retrieval ability and divergent thinking as well as the differences between these constructs (such as retrieving existing information versus generating novel combinations) require further exploration.

Given the theoretical relationship between retrieval, creative fluency, and originality it seems reasonable to think that they might be similarly predicted by the same cognitive abilities. Next, we summarize results reported for the relation between established cognitive abilities and retrieval and aspects of divergent thinking.

COGNITIVE ABILITIES THAT EXPLAIN VARIANCE IN RETRIEVAL ABILITY AND DIVERGENT THINKING

People differ in their ability to think divergently as well as in their ability to retrieve information from long-term memory. These individual differences seem to be partially explained by several cognitive abilities. For example, recent meta-analytical evidence shows not only that a factor accounting for general cognitive abilities is substantially related with overall creativity, but also that various second-order abilities such as fluid and crystallized intelligence, memory and learning related abilities, visual perception, cognitive and processing speed, and retrieval ability are linked to creativity (Serban, Kepes, Wang, & Baldwin, 2023). In addition, numerous studies report a link between retrieval ability and divergent thinking with working memory capacity (WMC) (Benedek, Jauk, Sommer, Arendasy, & Neubauer, 2014b; Lee & Therriault, 2013; Rosen & Engle, 1997; Unsworth, Spillers, & Brewer, 2011), secondary memory (Hedden, Lautenschlager, & Park, 2005; Madore, Jing, & Schacter, 2016), crystallized intelligence (Gc) (Unsworth, 2019; Weiss et al., 2021a), and mental speed (Forthmann et al., 2019; Miroshnik et al., 2023; Vock, Preckel, & Holling, 2011). These cognitive abilities have been previously discussed as crucial predictors of retrieval performance, as they are essential to activate, monitor, and suppress information during the actual retrieval process. A recent multivariate study has shown that these predictors explain 62% of the variance in a higher-order factor of retrieval ability (Goecke et al., 2024). Given the theoretical proximity of retrieval ability and divergent thinking, it seems promising to investigate whether these constructs predict divergent thinking to a similar extent. Next, we present earlier findings regarding the relation between divergent thinking and WMC, secondary memory, Gc, and mental speed.

THE PRESENT STUDY

For the present study, we aimed at examining two research aims. First, we addressed the question how creative fluency and originality are related with retrieval ability. Since creative fluency, expressed as a count variable, is conceptually close to retrieval we assume creative fluency to be highly correlated with retrieval ability. Furthermore, we assume that due to the positive manifold among ability measures originality is lower, but still substantially related with retrieval ability. Given prior results, we assume that model fit deteriorates substantially if creative fluency and retrieval fluency are assumed to correlate at unity.

Second, we examined the predictive power of cognitive abilities (working memory capacity, Gc, mental speed, and secondary memory) in the fluency factor(s) and originality. Based on previous findings, we expect that cognitive ability factors predict fluency and originality to a similar degree but might play a larger role in accounting for variance in fluency. Given that originality has been subsumed as a lower-order factor of retrieval ability in previous intelligence models, we further test if the residuals of retrieval and originality are meaningfully related after controlling for established cognitive abilities. A correlation of these residuals would indicate specificity in the link between both constructs. The research objectives of the present study were not preregistered.

PARTICIPANTS AND PROCEDURE

The study was promoted through multiple channels, including social media, mailing lists, flyers, and advertisements in newspapers and public transportation. Data collection occurred in two German cities. Participants needed to be fluent in German and aged between 18 and 45 years, considering that cognitive decline is observed more commonly in later stages of life (Hartshorne & Germine, 2015). Tasks and instructions were presented on 27″ computer screens using Inquisit 6 software (Inquisit 6, Version 6.5.2, 2022). The sequence of tasks was uniform across all participants. Each session had standardized on-screen instructions and was overseen by a trained proctor. The computerized test battery was conducted in group sessions, accommodating up to eight participants simultaneously. The entire test battery, with three breaks (two 5-min breaks and one 15-min break) including snacks and beverages, took about 5 h to complete. All participants provided written consent and received 55 Euros as compensation for their involvement. The study was conducted in adherence to the ethical guidelines outlined in the EU General Data Protection Regulation (GDPR), ensuring participant anonymity and receiving approval from a local ethics committee.

In total, N = 335 participants from the general population took part in our study1, from which N = 331 were eligible for statistical analysis, because n = 4 participants were excluded during data cleaning, because they either encountered major technical errors or had terminated the study early due to personal reasons. Please note that the sample size for the current study was determined based on a power analysis conducted for a multivariate study of retrieval ability (see also Goecke et al., 2024). The power analysis was based on the target measurement models of retrieval ability (single-factor vs. correlated factors model with three latent factors) and found that a sample size of N = 300 yields sufficient power (>.80, α = .01) for the targeted parameter estimates of these models (i.e., factor loadings). The sample of N = 331 were 65.4% female and ranged in age from 18–42 years (M _age = 26.4 years, SD _age = 5.4). N = 263 of the sample indicated holding at least a high school degree, and N = 85 indicated holding a college degree.

MEASURES

The focus of this manuscript is on the measurement of retrieval ability and divergent thinking. All constructs were measured by various tasks in a multivariate fashion as described in the following sections. Furthermore, cognitive abilities were assessed by measuring WMC, Gc, secondary memory, and mental speed. For all constructs displayed in table 1 measurement models based on aggregated items were estimated. The latent factors displayed acceptable to good reliability (ω_{RetrievalAbility} = .84; ω_{CreativeFluency} = .61; ω_Originality = .60).

DATA TREATMENT

Prior to statistical analysis, the data of interest for the current study were checked for outliers and implausible values. We investigated multivariate outliers across task scores in our sample based on the Mahalanobis distance (see also Meade & Craig, 2012). The Mahalanobis distance is the standardized distance of one data point from the mean of the multivariate distribution of several task scores. For our study, we considered participants fluency scores of three of the retrieval indicators and the scores of the three vocabulary tests to account for the verbal nature of most of the administered tasks. We excluded a total of n = 9 participants due to their Mahalanobis distances >20. Additionally, n = 2 more participants were removed as they were extreme outliers (>3 SDs) in the creativity task alternate uses, that is an often applied divergent thinking task. The final sample for statistical analysis thus consisted of N = 320 participants.

STATISTICAL ANALYSIS

All analyses were conducted with R (R Core Team, 2022) and we provide all materials necessary to reproduce the analyses in an online repository: https://osf.io/bt5zu/.

Confirmatory Factor Analysis (CFA) was carried out with the R package lavaan (Rosseel, 2012). Where possible, we used full information maximum likelihood estimation under the assumption of missing completely at random to combine missing data and parameter estimation in a single step (Schafer & Graham, 2002; Enders, 2010). Models based on continuous indicators are based on maximum likelihood estimator with robust standard errors (MLR). The following fit statistics were considered indicating good model fit: CFI (Comparative Fit Index) ≥ .95, RMSEA (Root Mean Square Error of Approximation) ≤ .06, and SRMR (Standardized Root Mean Square Residual) ≤ .08 (Hu & Bentler, 1999). For acceptable model fit these boundaries were used: CFI ≥ .90, RMSEA ≤ .08, and SRMR ≤ .10 (Bentler, 1990; Browne & Cudeck, 1992). We also computed dynamic fit indices (McNeish & Wolf, 2023), since fixed cutoffs for evaluating model fit are not a panacea for evaluating models (e.g., Groskurth, Bluemke, & Lechner, 2023). Therefore, we additionally evaluated model fit (CFI and RMSEA) from the empirical data against suggested dynamic cutoffs for “mediocre”, “fair”, and “close” model fit. Fair fit on CFI and RMSEA is recommended as a lower threshold for accepting models (McNeish & Wolf, 2023). Close fit is meant to indicate excellent fit of a model. Throughout the results, we transparently report both traditional and dynamic fit indices, but do not reject models based on single fit indices. We used McDonald's ω as indicator of factor saturation (McDonald, 1999; Raykov & Marcoulides, 2011). Factor saturation of a factor indicates how much variance is accounted for by a latent variable in all underlying indicators (Brunner, Nagy, & Wilhelm, 2012). Prior to modeling, indicator variables were standardized to account for differences in scales.

RETRIEVAL ABILITY, CREATIVE FLUENCY AND ORIGINALITY

First, we examined the relationship between retrieval ability, creative fluency, and originality. Both, retrieval ability and creative fluency are based on a count of correct answers and therefore express fluency of performance. The measurement models we examined are displayed in Figure 2. Traditional fit indices of model A showed acceptable fit, whereas the dynamic fit indices indicated that the model fit was below acceptability (i.e., <mediocre). Model A shows that retrieval ability and creative fluency are highly correlated (r = .88; SE = .054; p < .001). In contrast to that, retrieval ability and originality are related weaker but still substantially r = .59 (SE = .060; p < .001). Interestingly, creative fluency and originality—both aspects of divergent thinking—are still weaker related with one another (r = .43; SE = .099; p = .001) than both constructs are related with retrieval ability. Even though the correlation between retrieval ability and creative fluency is not at unity, setting this correlation to unity only leads to slight deterioration of fit (c.f., Model B in Figure 2). Given these results, retrieval ability and creative fluency can be subsumed below one fluency factor that entails ideational, expressional, word, or creative fluency. This factor is substantially related with originality (r = .60; SE = .061; p < .001).

ADDING COGNITIVE ABILITIES

Second, we aimed to examine to what degree general cognitive abilities account for variance in originality and fluency. Additionally, we also were interested in whether or not fluency and originality are related after controlling for cognitive abilities. Hence, we extended the model by a measurement model of cognitive abilities (for a detailed description of the measurement model see Goecke et al., 2024) that consists of a general factor (marked by the WMC tasks) and three nested factors capturing unique shares of variance of Gc, secondary memory, and mental speed respectively. We have chosen to model a bifactorial measurement model that is described as bifactor-(S-1) in the literature (Eid, Geiser, Koch, & Heene, 2017; Eid, Krumm, Koch, & Schulze, 2018). This model contains only three specific factors (S; in Figure 3: Speed, Secondary Memory, and Gc) and omits the fourth specific factor (WMC) in favor of a reference factor. Given the importance of WMC/fluid abilities for general cognitive abilities (e.g., Carroll, 1993; Kyllonen & Christal, 1990; Wilhelm et al., 2013) and given the methodological problems associated with full bifactor models (e.g., vanishing factors, irregular loading patterns, negative variances; Eid et al., 2017), this approach is warranted. Thismodel was then combined with Model B as displayed in Figure 2.

The structural equation model displayed in Figure 3 fitted the data well (dynamic fit: close fit). General cognitive abilities (g) predicted both retrieval ability (β = .60) and originality (β = .53) to a similar degree. Gc (β = .42 and β = .35, respectively) and secondary memory (β = .21 and β = .19, respectively) also predicted both factors fluency and originality. The regression weights from these nested factors indicate approximately similar proportions of variance accounted for by established cognitive abilities. Mental speed only predicted fluency significantly (β = .21), but did not play a significant role for originality (β = .08)—nevertheless, compared to the other predictors, both regression weights were substantially smaller. In total, 62.5% (R ² = .625) of variance in fluency was explained by the applied predictor space, and 45% (R ² = .450) of variance in originality was explained by the set of predictors. To investigate the unique relation between fluency and originality, after having controlled for the predictor space, we correlated their respective residuals (i.e., disturbance terms) with one another, which yielded a non-significant correlation (r = .11).

Please note that in Figure 3 all regression weights towards fluency and originality were freely estimated. To examine whether the amounts of explained variance in the factors of fluency and originality were statistically different from one another, we fitted an additional model where all regression weights for each predictor were set to be equal. Although the model still exhibited good fit (n = 320; χ²(405) = 795.69, CFI = .90, RMSEA = .06 with 90%-CI [.05; .06], SRMR = .07; dynamic fit: close fit), a formal model comparison revealed that the model displayed in Figure 3 (no constraints) fitted the data significantly better than Model B (∆χ2(4, N = 320) = 26.62, p < .001).

RETRIEVAL ABILITY AND CREATIVE FLUENCY

Retrieval ability was correlated highly with creative fluency. Both factors are based on tasks that express performance as count variables. The correlation between both abilities was close to unity and subsuming both abilities below one fluency factor hardly deteriorated fit. The correlation in the present data seems to be higher than relations previously reported either in multivariate individual studies or meta-analysis (see for example Miroshnik et al., 2023; Silvia et al., 2013).

However, previous theoretical and statistical models have stressed the closeness of both constructs that obviously lies in the method of performance appraisal if both are coded by a count variable. A general factor of fluency that subsumes retrieval ability and creative fluency is therefore in line with the proposed factor structure of Schneider and McGrew (2018) and Carroll (1993). In the creativity literature, such count variables are often used based on their time-efficiency and cost-effectiveness (e.g., Reiter-Palmon et al., 2019). Our study, however, shows that if indicators of creative fluency are analyzed based on simple count variables, they are highly related to the ability to retrieve information from long-term memory. Their extremely strong relationship, as well as the result that the model fit does not significantly diminish if both factors are collapsed to one factor even allows for the conclusion that these arguably two constructs (both displayed by count variables) are measuring the same underlying ability. Interestingly, creative fluency is more strongly related with retrieval ability than with originality. This implies that, creative fluency based on count variables has much less to do with originality than previously expected. Furthermore, the correlation between originality and retrieval ability is similar to the correlations that have been found hitherto (e.g., Miroshnik et al., 2023). This correlation indicates that retrieval ability and originality are related; however, they are far from being the same. The present result and the strength of the relation reported by Silvia et al. (2013) seem to match this. In fact, the reported correlations challenge assumptions that can be drawn from monotrait-monomethod and heterotrait-monomethod considerations (Campbell & Fiske, 1959). That is, given that creative fluency and originality are deemed two indicators of the same underlying construct, namely divergent thinking, they should be higher correlated with one another (monotrait-monomethod) than with another construct, that is not deemed part of the construct rationale of divergent thinking per se (e.g., retrieval ability: heterotrait-monomethod). In contrast to that assumption, our results rather endorse an interpretation of creative fluency being a plausible indicator of retrieval ability, whereas originality should be seen as something quite different.

Based on our findings, it is prudent to conclude that creative fluency indicators that are based on count variables are more akin to retrieval from long-term memory than they are to originality. These findings also provide further evidence for the existence of originality as a latent factor, if measured broadly, that can be distinguished from fluency factors and is meaningfully related to other abilities. In previous work, the existence of originality has been questioned, for example, due to limited factor variance of a nested originality factor (Weiss et al., 2021a) or due to redundancies and confounds with (creative) fluency (e.g., Batey, Chamorro-Premuzic, & Furnham, 2009; Preckel, Holling, & Wiese, 2006). Other studies show that originality, as indicated by various tasks using multiple items (e.g., combining objects task, creative consequences tasks, etc.), is indeed existent and discriminant to creative fluency (see, for example, Weiss et al., 2023). This inconclusiveness regarding originality can be traced back to insecurities regarding the many faces this construct can be measured. Taken together, our present results rather stress the importance of originality as a distinct factor (if measured broadly based on indicators asking for one very original answer, but also including indicators that average the originality across multiple answers) and we suggest—despite their time and cost efficiency—to refrain from using count variables, but rather use originality indicators.

FLUENCY AND ORIGINALITY BEYOND COGNITIVE ABILITIES

Prior research endorses the view that relations between indicators of fluency (e.g., retrieval indicators) are not simply due to these indicators being a linear combination of various established cognitive abilities (Goecke et al., 2024). Similarly, divergent thinking has been related to cognitive abilities. However, most studies are restricted to measures of fluid and/or crystallized intelligence and might thus not exhaust the predictive power of established intelligence factors (e.g., Gerwig et al., 2021). Given the conceptual proximity of retrieval ability and originality, we expected that cognitive abilities that have been shown to explain substantial variance in retrieval ability (WMC, Gc, secondary memory, and mental speed) also and to a similar size explain variance in divergent thinking. Explicitly, we found a general factor (referenced by WMC) of intelligence explaining the most variance in originality, while nested Gc was the second-highest predictor before secondary memory and speed (that was unrelated with originality). Recent meta-analytical findings on the relationship between creativity and various Stratum II abilities (c.f., Three-Stratum Theory of Cognitive abilities; Carroll, 1993) support our findings and report correlations of a similar magnitude. In sum, we found that these constructs explained 47% of the variance in divergent thinking, compared to 63% of the variance in fluency (indicators of retrieval ability and creative fluency). Previous studies have related fluency (as an indicator of retrieval ability and creative fluency) as well as originality with these cognitive abilities and found similar predictive power. Given the overlap between fluency and originality and previous results where divergent thinking/originality assessments were often contaminated with fluency resulting in increasing correlations between the two (Forthmann, Szardenings, & Holling, 2020; Hocevar, 1979) the assumptions that originality is ordered among fluency inherently makes sense. However, most studies have not included the variety of cognitive abilities, fluency indicators and originality to go further and test if fluency and originality really share communalities beyond cognitive abilities. In our study, we found that residuals of fluency and originality are not meaningfully related, once established cognitive abilities are controlled for. This implies that the two constructs have nothing in common beyond the variance that is explained by Gc, WMC, and secondary memory. Based on this finding we would argue that (a) originality and the fluently retrieved information from long-term memory are much less related than previously suggested; and (b) the results question theoretical models that subsume originality as a part of retrieval ability (e.g., Schneider & McGrew, 2018). Therefore, we would argue that originality is a unique ability that is related with other cognitive abilities, but not ordered among one of them.

LIMITATIONS AND DESIDERATA FOR FUTURE RESEARCH

The current study included a broad variety of indicators and used a large community sample. Nevertheless, there are a few limitations that have to be kept in mind when interpreting the results. First, creative fluency was only assessed based on two tasks (each including four items). More tasks are desirable to allow better abstraction from peculiarities of specific indicators. Second, the tasks that were used to measure divergent thinking were all based on verbal indicators, whereas retrieval ability also included a figural indicator. Future studies should cover a broader range of task content to obtain better construct coverage. Third, the scoring procedures applied in the creative fluency as well as the originality tasks might challenge the current results: different scoring procedures might have led to (slightly) different results. For example, if only truly novel answers were scored as correct responses in creative fluency tasks and if all answers that seem to display a pure retrieval from long-term memory were dismissed (Benedek et al., 2014a, 2014b; Gilhooly, Fioratou, Anthony, & Wynn, 2007; Silvia, Nusbaum, & Beaty, 2017), this scoring could have affected the relationship between creative fluency and retrieval ability as well as originality. In such a case, the correlation between creative fluency and retrieval ability might be lower than reported in this study. We thus want to stress that future research might benefit from studying such moderating effects. This also applies to various scoring procedures that could be applied to score originality. In our current study originality was scored by rating each and every answer. However, a recent quantitative review of human scoring methods shows more than 100 different instructions and scoring procedures of divergent thinking tasks (Saretzki et al., 2024). Future studies should systematically investigate how different scoring approaches of creative fluency tasks and originality tasks and aggregation methods in scorings are related to one another and how they affect relations with other cognitive abilities. Such a study should also include variants of computerized scorings that are based on large language models (e.g., Organisciak, Acar, Dumas, & Berthiaume, 2023).

CONCLUSION

In sum, we found that retrieval ability and creative fluency were highly related to one another. Originality was related to both, however, to a much smaller degree. We conclude that the usage of count variables in creativity research does rather capture retrieval. Interestingly, originality and the ability to fluently retrieve information from long-term memory were not related to another beyond working memory capacity, crystallized intelligence, and secondary memory. These empirical findings stress the unique position of originality and its distinctiveness of fluency and question several theoretical intelligence models that place originality among factors of the ability to retrieve information. We conclude that the placement of originality among retrieval ability should be rethought.