Predictors of training effi cacy during n-back training Andrea Vranić 1* , Marina Martinčević 1 and Vedran Prpić 2 1 Department of Psychology, Faculty of Humanities and Social Sciences, University of Zagreb, Croatia 2 Ina d.d., Human resources, Croatia Abstract: Numerous studies have investigated the effi cacy of various cognitive trainings, with working memory being the most often trained cognitive aspect. In this regard, executive aspects of working memory have received the most attention, with updating training being vastly explored. In this study, we aimed to examine the diff erential contribution of some individual characteristics to the effi cacy of updating training using a well-established n-back training paradigm. More specifi cally, we examined the contribution of fl uid reasoning (gf), and personality (neuroticism, conscientiousness) to training effi cacy. Participants (N = 47) took part in a 15-session, dual n-back training, spread over 4 weeks. They were pretested for fl uid reasoning (CFT-3), personality (IPIP-100), and performed the initial testing on the OSP AN task. OSP AN was measured in three additional measurement points (after 5th, 10th, 15th session). The data was analyzed within the multilevel modeling approach. Initial hypotheses were partly confi rmed, in that: 1) training was effi cient in terms of OSPAN score, which grew linearly over time and the trajectory was similar between participants, 2) although the growth was similar for all participants, diff erences were found in intercepts, and 3) these diff erences could be partly explained by diff erences in fl uid reasoning, but not with personality traits of conscientiousness and neuroticism. Keywords: working memory training, working memory capacity, dual n-back, conscientiousness, neuroticism Napovedniki u činkovitosti treninga v razli čnih fazah treninga n-nazaj Andrea Vranić 1* , Marina Martinčević 1 in Vedran Prpić 2 1 Department of Psychology, Faculty of Humanities and Social Sciences, University of Zagreb, Croatia 2 Ina d.d., Human resources, Croatia Povzetek: Številne raziskave so preučevale učinkovitost različnih kognitivnih treningov, pri čemer je bil delovni spomin najpogosteje trenirana kognitivna funkcija. Največ pozornosti so bili deležni izvršilni vidiki delovnega spomina in pri tem je bilo še posebej temeljito raziskano posodabljanje informacij. V pričujoči razisk av i smo z up or ab o dobro uveljavlje ne pa r a d ig me, t j. pa r a d ig me n-nazaj, želeli preučiti, kakšen je prispevek fl uidne inteligentnosti oz. sklepanja (gf) in osebnosti (nevroticizma in vestnosti) k učinkovitosti treninga posodabljanja informacij. Udeleženci (N = 47) so v 15 seansah skozi 4 tedne izvajali trening z nalogo dvojnega n-nazaj. Pred treningom smo s testom CFT-3 izmerili njihovo sposobnost fl uidnega sklepanja, s testom IPIP-100 njihovo osebnost, reševali pa so tudi nalogo obsega operativnega spomina (t. i. nalogo OSPAN). OSPAN smo izmerili še v treh dodatnih časovnih točkah (po 5., 10. in 15. seansi treninga). Podatke smo analizirali s postopki večnivojskega modeliranja. Prvotna hipoteza je bila delno potrjena: (1) t rening je bil učinkovit, saj je dosežek na nalogi OSPAN s časom linearno naraščal, naraščanje pa je bilo podobno pri različnih udeležencih; (2) čeprav je bila funkcija naraščanja dosežka pri različnih udeležencih podobna, smo našli medosebne razlike v presečiščih; (3) te razlike smo lahko delno pojasnili z medosebnimi razlikami v sposobnosti fl uidnega sklepanja, ne pa tudi z razlikami v osebnostnih potezah vestnosti in nevroticizma. Ključne besede: trening delovnega spomina, obseg delovnega spomina, dvojni n-nazaj, vestnost, nevroticizem Psihološka obzorja / Horizons of Psychology, 30, 129–137 (2021) © Društvo psihologov Slovenije, ISSN 2350-5141 Znanstveni empiričnoraziskovalni prispevek / Scientifi c empirical article DOI: 10.20419/2021.30.530 CC: 2343 UDK: 159.953 * Naslov/Address: Andrea Vranic, Department of Psychology, Faculty of Humanities and Social Sciences, University of Zagreb, Ivana Lučića 3, 10000 Zagreb, Croatia, e-mail: avranic@ff zg.hr Članek je licenciran pod pogoji Creative Commons Attribution 4.0 International licence. (CC-BY licenca). The article is licensed under a Creative Commons Attribution 4.0 International License (CC-BY license). 130 W orking memory (WM) is a fl exible multi-component mental workspace conceptualized via complementary theori es f or o v er fi ve decades. Converging assumption of these theories defi ne WM as a system of multiple capacity- limited domain-specifi c cognitive mechanisms, regulated by an executive unit. What was once viewed as a single central executive component involved in the integration and storage of multimodal information and simultaneous supervision of control processes, is now conceptualized as a range of executive functions (EF), such as attentional control or task- switching (Logie, 2011). EF are a set of top-down cognitive processes indispens- able for all aspects of cognitive functioning, spanning from basic (cognitive control, inhibition, fl exibility) to higher order processes, such as planning, organization or reasoning (Diamond, 2013). Probably the most prominent theoretical and empirical framework in this fi eld is the tripartite model, which distinguishes three rudimentary EF: mental set shifting, information updating and monitoring, and inhibi- tion of prepotent responses (Miyake et al., 2000). Updating requires monitoring and coding of incoming information relevant to the task, and appropriately replacing old, no longer relevant information held in WM with newer, more relevant information (Morris & Jones, 1990; Miyake et al., 2000). Shifting refers to fl exibility in altering back and forth between multiple tasks, operations, or mental sets (Altmann & Gray, 2008). Inhibition is ability to deliberately inhibit a dominant, automatic, or prepotent response when necessary (Miyake et al., 2000). This three EF play an important role in memory and learning. Even though all three EF are consider- ed of major importance to the functioning of WM, some authors emphasize updating to be central to WM (Ecker et al., 2010). This notion is further corroborated by the fi ndings showing signifi cant correlations of complex span tasks (WM tasks) and updating (Schmiedek et al., 2009; 2014). Updating training Given the importance of WM in everyday life, a large number of studies has addressed the question of whether WM can be enhanced through cognitive exercise. Cognitive training is based on the mounting evidence of brain plasticity, not only at an early age, but also much later in life (Pauwels et al., 2018). Intensive learning of new or non-routinized skills, typical for cognitive training, is thus an experiential setting e s t a b l i s h e d t o p r o m o t e b r a i n p l a s t i c i t y . I n d e e d , i m a g i n g s tu di e s s h o w a l t e r e d b r a in a c ti vi ty , a s w e l l a s s tru c tur a l grey and white mater changes, related to improvement and refi nement in performance on various (cognitive) tasks, in young and old alike (Zatorre et al., 2012). Overall, it seems t h a t W M t r a i n i n g r e s u l t s i n q u a n t i t a t i v e s h i f t s i n b r a i n activity (Buschkuehl et al., 2012). Co n v en ti o nal WM trainin g co ns i s ts o f re gul ar perf o r - mance on a complex WM task. A common feature of such a training is the adaptive task tailored to challenge parti- cipant’ s level of ability. The most frequently used updating training task is the n-back, which requires participants to indicate whether the currently presented stimulus matches the one presented n trials previously. N-back was fi rst introduced by Kirchner (1958) in an investigation of age diff erences in short-term memory, and although studies have criticized it as a measure of WM (Kane et al., 2007), its face validity – short-term maintenance and constant updating of information required to perform the task – has helped it become the signature paradigm of WM updating training. A recent meta-analysis showed that apart from benefi ting young adults’ updating ability, n-back training can benefi t other cognitive functions, such as fl uid reasoning, and the eff ects can last up to 18 months (Au et al., 2015; but for older adults see Lampit et al., 2014). In this study we have used OSPAN as a WM span task and a proxy to training effi cacy. What is the relation of n-back to OSPAN scores? N-back has face validity as a WM task since it requires continuous maintaining, updating and processing of information (Gajewski et al., 2018). Dual n-back requires a skill to manage two n-back tasks simultaneously, which is a common requirement of tasks tapping Gf. Studies which have employed OSPAN as a measure of WM training effi cacy suggest a positive interrelation of OSPAN and dual n-back training effi c a c y . F o r e x a m p l e , M a t y s i a k a n d c o l l e a g u e s (2019) have found that that among a number of variables only baseline OSPAN performance was found to be a signifi cant predictor of the n-back result at the fi rst day of training and a moderator of the whole training course. Individual predictors of training benefi ts Among potential moderators of transfer eff ects studies a g r e e o n a g e , w i t h y o u n g e r p a r t i c i p a n t s h a v i n g l a r g e r improvements compared to older ones (e.g. Melby-Lervåg & Hulme, 2013). However, a moderating role of training dose is disputed. While no eff ect of dosage is found when this variable is dichotomized based on a median-split (Melby- Lervåg & Hulme, 2013; Soveri et al., 2017), authors which have used continuous moderation fi nd that higher dose training yields larger mean eff ects (Schwaighofer et al., 2015). In sum, training dose varies substantially between studies and studying its eff ects should not be easily abandoned. An a ft e r - a n a l y s i s o f s tu d i e s i n v e s t i g a t i n g t h e e ffi cacy of updating training revealed diff erential changes among trained participants; positively challenged participants have demonstrated large gains, while others have shown no improvement with some showing signs of regressing (Jaeggi et al., 2011). It might be that updating trainings produce such mixed results for they do not single out the factors which might catalyze its effi cacy (Wiemers et al., 2019). Identifying factors which contribute to training benefi ts has become an important new research avenue. Our study follows this line of research in dealing with the impact of individual diff erences on training outcomes (e.g. Colquitt et al., 2000; Studer-Luethi et al., 2012a, 2012b). Suggesting an interaction of aptitude and treatment, a straightforward candidate for moderation of pretest-posttest scores on various cognitive tasks are general individual diff erences in cognitive ability. Indeed, evidence suggests that individual diff erences in training-related gains can be partly explained by initial cognitive resources; magnifi - cation hypothesis suggests that individuals with good A. Vrani ć, M. Martin čevi ć and V . Prpi ć 131 Predictors of training e ffi cacy during n-back training default abilities will profi t more leading to magnifi cation of performance (Lövden et al., 2012). This so-called “rich get richer” eff ect speaks of initial diff erences in WM and Gf as the predictor of training-related gains and transfer eff ects (Redick et al. 2015; von Bastian & Oberauer, 2014). On the other hand, compensatory account assumes that people with lower initial abilities can benefi t more from cognitive training because they have more room for improvement, i.e. training can aid more to individuals with lower cognitive abilities (Titz & Karbach, 2014). Other than cognition-related factors, specifi c personality characteristics have been identifi ed as a potential contributor to training success. Studies consistently show that anxiety, emotional instability, and depression are negative predictors o f t r a i n i n g s u c c e s s , a n d t h e s e p e r s o n a l i ty c h a r a c t e r i s t i c s usually relate to neuroticism (Naquin & Holton, 2002; Studer- Luethi et al., 2012a). Individuals with high neuroticism often have the least benefi t from trainings. The unsubstantiated explanation off e r e d b y s t u d i e s i n v e s t i g a t i n g t h e r e l a t i o n of neuroticism and training outcomes is in line with the attentional control theory which postulates that anxiety (e.g., worry) exerts additional cognitive load and leaves insuffi cient resources for further general processing (Eysenck et al., 2007; Grimley et al., 2008). Furthermore, a comparison of after-training performance between high and low demanding training conditions (single vs. dual n-back) shows lower gains for participants with higher neuroticism in more complex dual n-back condition ( e.g. Studer -Luethi et al., 20 12a ). In simpler tasks, higher initial arousal level sustains vigilance and attention in individuals with high neuroticism, while individuals with lower arousal level seemed to have been understimulated. F u r t h e r m o r e , e v i d e n c e i s f o u n d w h i c h s u g g e s t s conscientiousness to be among the traits aff ecting training outcomes (Colquitt et al., 2000). Since conscientious indivi- duals are persistent and self-disciplined, conscientiousness is usually related to positive training outcomes (e.g., Tziner et al., 2007; Stueder-Lothi et al., 2012b). Since meta-analysis does not support the association between consciousness and skill acquisition, it is often assumed that motivation plays a major role in the relation of conscientiousness and training outcome (Colquitt et al., 2000). Still, there seems to be no s t r a i g h t f o r w a r d i n t e r p r e t a t i o n o f t h i s r e l a t i o n , a n d t h e interaction of conscientiousness and training interventions is likely to be somewhat complex. The aim of this study In the current study, we investigated whether individual diff erences in fl uid reasoning, conscientiousness and neuroticism, previously found related to cognitive outcomes ( e. g. , E ysenck et al. , 2007) , might contribute to cognitive training effi cacy. More specifi cally, we aimed to investigate the contribution of these traits in diff eren t p hases o f th e training program. T raining dose, as already mentioned, is a potentially important moderator of training outcomes, and studying variables with a diff erential eff ect in diff erent phases of training or skill acquisition might shed additional l i g h t o n t h e c o m p l e x i n t e rp l a y o f t r a i n i n g d o s e a n d i t s outcomes. W e employed a training procedure ( same training task and format) previously validated on a sample of psychology students (Tkalčević, 20 13; Tkalčević & Vranić, 20 13 ). The effi cacy was investigated on the measures of fl uid reasoning ( C F T - 3 ) , a t t e n t i o n ( d 2 t e s t , B r i c k e n k a m p , 1 9 9 9 ) a n d spatial memory (TPP , Vranić, 201 1 ) and it showed posttest improvements for all measures which were maintained at the 6 months follow-up, although to a somewhat lesser extent. W e h ypo th es iz ed tha t fl ui d r eas o nin g a b ili ty will ha v e a benefi ciary eff ect on the training outcome, providing a linear, “rich-get-richer” increase throughout the training (Lövden et al., 20 12). Furthermore, we hypothesized that neuroticism will have a larger (negative) impact in the initial training sta- ges, while its eff ect should decrease over time. It seems that more experience with the situation and task requirements, as w ell as the f eedback about progress – all o f which is provided by repetitive exercise in adaptive training – reduces the infl u e n c e w h i c h n e ur o t i c i s m m i g h t e x e rt o n tr a i n i n g o u t c o m e s ( e . g . E y s e n c k e t a l . , 2 0 0 7 ) . Further along, we wanted to investigate whether conscientiousness will exhibit a diff erential eff ect on training effi cacy. Method Participants Undergraduate psychology students at the University of Zagreb participated in this study. A total of 47 students with an average age of 19 years (range 18–23 years) were included. M o s t p a r t i c i p a n t s w e r e f e m a l e ( 8 9 % ) . A l l p a r t i c i p a n t s provided their written informed consent to participate in the study. Training task Dual n-back task (Jaeggi et al., 2010). An adaptive updating task (described in Jaeggi et al., 2010). The task consists of two simultaneous n-back tasks (spatial and verbal). In a spatial task a single blue square is sequentially presented at one of eight diff e r e n t s c r e e n l o c a t i o n s . I n a verbal variant, a one among eight diff e r e n t l e t t e r s i s presented through headphones. The task requires a response (key press) only when the currently presented stimulus (square location and/or letter) are the same as the ones n positions back in the sequence. Each session included 20 blocks consisting of 20+n trials (total of 20 min). Feedback is given after each block and the n level changes depending on participants ’ performance : If participant’s performance was above 90% accurate, n-level increased for n = 1, and if performance was less than 75% accurate, n-level decreased for n = 1. Tests Operational span task (OSP AN; Unsworth et al., 2005). OSPAN is a WM task in which participants try to remember sequentially presented letters in their correct order while simultaneously solving simple math equations. A feedback is provided on the number of correctly recalled letters, as well 132 as the percentage of correctly solved mathematical equations. Participants had to be correct in at least 85% of the equations. The task consists of sets of 3-7 equations and letters, and a set of each length is presented three times (75 sets in total). The order of sets was randomized across participants. OSPAN result represented the sum of the letters in the accurately recall ed sequence . F or examp l e, if a parti ci pan t correctl y recalls 2 letters in a 2-letter sequence, 3 letters in a 3-letter sequence and 3 letters in a 4-letter sequence, his result is 2 + 3 + 0 = 5. Cattell’s Culture Fair Intelligence Test, Scale 3 (CFT- 3). CFT-3 is a non-verbal test that measures fl uid reasoning, excluding the infl uence of verbal ability, culture and educational attainment. The test consists of 4 subtests: 1) Series (13 tasks), 2) Classifi cations (14 tasks), 3) Matrices (13 tasks), and 4) Conditions (10 tasks). It takes approximately 15min to complete. Based on the answers to all subtests, a single total score is formed that can be directly transformed into the intelligence quotient score (IQ score). In this sample, average IQ score was 126.6 (SD = 12.40; range: 103–157). International Personality Item Pool - 100 (IPIP-100, Goldberg, 1999). IPIP-100 is a cross-cultural personality questionnaire constructed according to the Big-Five model. It consists of 100 items represented by short statements and the task is to indicate, on a scale from 1 (“Very Inaccurate”) to 5 (“Very Accurate”), how much each statement describes the participant himself. IPIP-100 covers fi ve personality traits: extraversion, agreeableness, emotional stability, conscientiousness, and intellect. Each trait is measured by 20 statements. In this research, we used only items referring to emotional stability and conscientiousness. Cronbach alpha in this study showed good internal consistency (emotional stability α = 0.94; conscientiousness α = 0.93). Procedure Participants took part in the dual n-back WM training. The training consisted of 15 individual sessions of 20 min- utes, extended over every other day for 4 weeks. Prior to the training participants fi lled in the IPIP- 1 00, CFT - 3, and have completed the OSPAN task. During the training, three additional measurement points were conducted at which the performance in OSPAN task was measured (after 5 th , 10 th , and the last (15 th ) training session). Participants were given class credits for their participation. Ethical approval was obtained from the faculty ethics committee. Statistical analysis We analysed the data within the multilevel modelling a p p r o a c h . T h i s a n a l y s i s i s a p p r o p r i a t e f o r h i e r a r c h i c a l l y nested data in which a lower level unit of analysis is nested within a higher level of analysis. Our data had a two-level structure: repeated measure- ments over time (level-1) and participants (level-2). Measurement occasion was introduced as a level- 1 predic- tor, and participant’ s CFT -3 score, level of neuroticism and conscientiousness were introduced as level-2 predictors. All predictors were centred by subtracting the mean from each value of the measured variable. As the fi rst step in the analysis, the intercept-only model was computed (model 1). This model gives the estimate of the intraclass correlation ρ which indicates the proportion of the variance explained by the grouping structure in the popula- tion (Hox, 2010). As a second step, time variable was added as a predictor in the model. In this step we tested models with linear (model 2a) and quadratic change (model 2b) of the dependent variable over the time. As a third step, the slope of the time variable was allowed to vary across individuals (model 3). In the last step, we entered the predictor variables at the second level (model 4). For parameter estimation, we used the maximum likelihood estimation (ML). Models were nested, therefore chi-square diff erence test based on the deviance statistic was used to compare models. All hypotheses were tested at the 5% alpha error rate. Results Descriptive statistic and the correlation matrix of dependent variables and the predictors are given in Table 1. Moderate to large positive correlations were found for OSPAN results at diff erent time points, and moderate positive correlation was found for CFT-3 score and OSPAN score at the last time point. Other correlations were non-signifi cant. Mean personality scores fi tted the mid-range values. A peculiarity of the sample is a rather high Gf score (95 th percentile). The results of MLM analysis are shown in Table 2. The fi rst model was used to calculate intraclass correlation (ICC) that shows the proportion of between-persons variance in the total variance. In our sample, the ICC was 0.44 indicating that there is within-person change over time ( 56% of total Table 1 Descriptive data for OSPAN (at four measurement points), CFT-3 score, neuroticism and consciousness for (N = 47) Variable MS D1. 2. 3. 4. 5. 6. 7. 1. t 1 OSPAN 41.96 14.503 2. t 2 OSPAN 46.17 14.344 .62 ** 3. t 3 OSPAN 50.45 14.336 .47 ** .46 ** 4. t 4 OSPAN 55.40 13.322 .67 ** .42 ** .63 ** 5. CFT-3 126.55 12.395 .25 .22 .21 .32 * 6. Neuroticism 56.87 13.760 –.21 –.26 –.04 –.11 –.08 7. Consciousness 63.68 13.239 –.30 * –.27 –.07 –.16 –.09 –.14 * p < .05; ** p < .01 A. Vrani ć, M. Martin čevi ć and V . Prpi ć 133 Predictors of training e ffi cacy during n-back training variance), but it is also worth grouping by persons to explain the variation in change between-persons (44% of total variance; Allerhand, 2016). We tested linear and nonlinear eff ects separately in order to defi ne the shape of the growth trajectory in model 2. An examination of the fi xed eff ects of model 2a in Table 2 indicates that there is a signifi cant linear relationship between time and OSP AN scores. Compared to model 1, this model showed better fi t to data, χ 2 (1) = 44.57, p < 0.01. Model 2b w i t h q u a d r a t i c n o n l i n e a r t r e n d s h o w e d n o n - s i g n i fi cant change from Model 2a, χ 2 (1 ) < 1, p = 0.79, and quadratic eff ect of time was not a signifi cant predictor. Therefore, in all presented models trajectories of time-related OSPAN score were modelled through linear change. In the next step we tested whether model with random slopes (model 3) fi ts the data better than the model with the fi xed slope. Although, a somewhat better fi t was found for mode l 3 , the results o f chi-square sho w ed non-si gnifi cant change from model 2a, χ 2 (2) < 1, p = 0.88, suggesting the same pattern of change between individuals OSP AN scores over the time (Figure 1). In our next step we added time-invariant predictors (level 2 predictors; model 4a). In this model we tried to explain between-person variation in the dependent variable (the between-person diff erences in the intercept). This model showed signifi cantly better fi t c o m p a r e d t o m o d e l 2 a , χ 2 (3) = 9.59, p < 0.05, but the only signifi cant predictor was CFT-3. Adding the level-2 predictors, we explained 22.3% of b e tw e e n - s u b j e c t v ari an c e c o m p ar e d t o m o d e l 2 th a t w a s used here as a baseline model. In order to enable the interpretation of intercept as the baseline measurement, in model 4b the initial testing (1 st measurement point) was set as zero. This model predicts the initial OSPAN score of 41.8, which increases by 4.5 points at each succeeding measurement point. With each scale point higher on the CFT-3 test, OSPAN score increased by 0.3. The results of this analysis suggest that OSPAN score grew linearly over time and the trajectory is similar between participants. Although the growth is similar for all participants, there are diff erences in intercepts, which might be partly exp lained by the diff erences in fl uid reasoning, but not with the personality traits of conscientiousness and neuroticism. Discussion This study was set with the aim to further elucidate the diff erential contribution of some individual characteristics to the benefi ts of cognitive training. More specifi cally, we o p t e d t o i n v e s t i g a t e t h e c o n t r i b u t i o n o f fl uid reasoning, conscientiousness and neuroticism to cognitive training effi c a c y i n t h e d i ff e r e n t p h a s e s o f t h e 1 5 - s e s s i o n W M updating training. Based on the literature, we hypothesized that fl uid reasoning will have a linear benefi ciary eff ect in diff erent phases of training, in line with the rich-get-richer eff ect (Lövden et al, 2012), and that neuroticism will have a larger impact in the initial training stages. Due to ambiguity of fi n d in g s o n r e l a ti o n o f c o n s c i e n ti o u s n e s s an d tr a inin g outcomes (e.g., Eysenck et al., 2007), no directional prediction was set in this case. Table 2 Change of OSPAN scores over time as predicted by CFT-3 and personality traits (N = 47 participants; 188 observations) Fixed eff ects Model 1 Model 2a Model 2b Model 3 Model 4a Model 4b Predictor Coeffi cient (SE) Coeffi cient (SE) Coeffi cient (SE) Coeffi cient (SE) Coeffi cient (SE) Coeffi cient (SE) Intercept 48.5 (1.66) ** 48.5 (1.66) ** 48.3 (1.88) ** 48.5 (1.66) ** 48.5 (1.51) ** 41.8 (1.78) ** Time 4.5 (0.62) ** 4.5 (0.62) ** 4.5 (0.63) ** 4.5 (0.62) ** 4.5 (0.62) ** Time 2 0.2 (0.69) CFT-3 0.3 (0.12) * 0.3 (0.12) * Neuroticism –0.2 (0.12) –0.2 (0.12) Consciousness –0.2 (0.11) –0.2 (0.11) Random eff ects σ 2 e 123.41 89.11 89.11 87.83 89.23 89.23 σ 2 u0 98.01 106.31 106.32 105.76 82.59 82.59 Deviance 1504.80 1460.22 1460.15 1459.97 1450.64 1450.64 * p < .05; ** p < .01 Figure 1 Individual changes in OSPAN scores during training at four measurement points (N = 47) 0 1 Time 2 3 20 40 OSPAN score 60 134 Our data was modelled using a multilevel approach, with repeated measurements over time (4 time points; level 1) and participants (level 2). We used the OSPAN task (Unsworth et al., 2005), a complex WM span task, as a criterion of training effi cacy in eac h o f the three phases, f o ll o wing the ini tial testing . Our results partly support the set hypothesis. We found that OSPAN scores show a comparable linear growth across diff erent training phases, and that the diff erences found in intercepts can be partly explained by initial diff erences in fl ui d r eas o nin g . Our h ypo th es i s r e g ar din g th e diff erential contribution of conscientiousness and neuroticism to the training effi c a c y w e r e n o t c o n fi rmed. Although not as o v erwhe lming in terms o f expectati ons met, our fi ndings cast some interesting light on the widely discussed relation of fl uid reasoning, n-back task and span tasks and WM ability. We will address this topic in later sections of the paper, and we will fi rst comment on the results regarding personality traits. Personality as a predictor of training outcomes Insofar, only a handful of studies have investigated the relativ e contribution o f non-cognitiv e factors, such as personality, to training eff ectiveness and outcomes, although there is an upward trend in the number of such studies (e.g. Colquitt et al., 2000; Jaeggi et al., 2014; Studer-Luethi et al., 2012a, 2012b). The importance of these studies is that they could enable the design of individualized programs, better suited for each individual, which in turn could boost the effi cacy of such personalized procedures. Our expectation of neuroticism to be relevant in determining the (in)effi cacy o f the training was based on the fi ndings o f anxiety and depressive symptoms being negatively related to training effi cacy (Backman et al., 1996; Studer-Luethi et al., 2012a). Emotional instability and anxious behavior are often used to describe individuals with high neuroticism scores. Almost contrary to this, individuals high on conscientiousness, which are described as reliable, persistent and diligent, could be expected to profi t from the training, although the motivation might play a critical role in determining the direction of this relation. Studer-Luethi and colleagues (2012a; 2012b) have found a detrimental eff ect of neuroticism in dual n-back training, such as the one employed in this study. The same study, h o w e v e r , f o u n d t h a t p a r t i c i p a n t s w i t h h i g h n e u r o t i c i s m exhibited higher gains in fl uid reasoning measures after the single n-back training. These results have off ered a specula- t i v e a c c o u n t o f t h e r e l a t i o n o f n e u r o t i c i s m a n d t r a i n i n g effi cacy, in line with the processing effi ciency theory (Eysenck & Calvo, 1992). It is suggested that anxiety – and neuroticism, by the same token – can be a useful feature in solving simple tasks due to the increased activation that helps sustain vigilance. But when the task is more demanding, the same feature imposes cognitive load on further processing, and in turn reduces processing capacity within WM. Anxiety is often present in novel situations, which can provoke lower self-effi cacy and less effi cient resource mobilization, while repetitive exercise within adaptive training provides opportunity to get acquainted with the task setting which consequently weakens the impact of neuroticism. However, in our study neuroticism did not have a diff erential eff ect on dual n-back training effi cacy. Conscientious individuals are usually motivated to improve and excel in their skills and performance (Komarraju & Karau, 2005), which leads to the assumption that they will be equally eager and hard-working in the training c o n d i t i o n s . P r e v i o u s s t u d i e s h a v e p a r t l y c o n fi rmed this notion, with participants high in conscientiousness having greater training success in the single n-back compared to dual n-back condition (Studer-Luethi at al., 2012). Authors argue that these individuals might have preferred lower task complexity because it provides higher chances for excellent performance. Following on that, dual n-back condition, such as the one in our study, could have reduced their resources via increased self-attention, and the tendency of conscientious individuals to be self-deceptive (Martocchio & Judge, 1997; Studer-Luethi et al., 2012b). However, results of our study which regard complex dual n-back condition did not reveal a signifi cant contribution of conscientiousness to the training outcomes, neither in positive, nor in negative manner. Although relatively small, our sample was twice the size of the dual n-back group in the study of Studer-Luethi and colleagues (2012a), and it could be considered as m o re ro b ust. H o w e v er , gi v en a body o f li tera ture o n th e diff erentially impaired cognitive processing in neuroticism, much research is needed to give a fi rm conclusion on the interplay of neuroticism and cognitive training outcomes. By the same token, the relation of conscientiousness and training outcomes needs a more systematic research approach, with varying levels of task complexity. Fluid reasoning as a predictor of training outcomes W e now come to the discussion of the fi ndings on the interrelation of cognitive variables within this study: Gf, n- back and OSPAN. Fluid reasoning can briefl y be defi ned as the ability of successful reasoning with novel problems. It vastly relies on working memory capacity (WMC; e.g., Kyllonen & Christal, 1990). WMC accounts for a signifi cant portion of the variance of general cognitive ability . WMC tasks – the so-called span tasks, such as OSPAN – have enjoyed strong support for being stable, and with a minimal contribution of error (e.g. Conway et al., 2005). Performance on WM span tasks correlates with a number of higher order cognitive skills, such as language comprehension (Daneman & Carpenter, 1983), reasoning (e.g. Barrouillet, 1996), and complex-task learning (Kyllonen & Stephens, 1990). Although evidence suggests that n-back and WM span are weakly correlated (Gajewski et al., 2018), in our study n-back training has resulted in gains on OSPAN task. Moreover, OSPAN scores at diff erent phases o f training w ere f ound related to initial scores in fl uid reasoning (Gf). Insofar, these fi ndings provide support for the notion that: 1) WM shares considerable variance with Gf, and 2) even though weakly correlated as expected, n-back and OSPAN might share some processing components that, in turn, could enable transfer A. Vrani ć, M. Martin čevi ć and V . Prpi ć 135 Predictors of training e ffi cacy during n-back training between these tasks. In light of the recent critiques of n-back as a measure of WM, our results provide support for some common ground of n-back and WMC. Thus, future studies should not discard n-back as a WM measure, rather focus on the specifi c context in which WM updating might prove benefi cial for a more general cognitive functioning. Also, keeping in mind the relatively high initial Gf score of our participants, further support for the interrelation of n-back and span tasks is needed from the samples with lower mean scores. P r e v i o u s s t u d i e s h a v e f o u n d c o n t r a r y fi ndings about the relationship between initial abilities and training gains which can be explained by the two opposite eff ects: “rich get richer” account (Lövdén et al., 2012) and compensatory eff ect (Karbach and Spengler, 2012). The fi ndings of our study do not confi rm the existence of any of those eff ects. Although fl uid reasoning is associated with WMC, it does not explain changes in the training gain. It is possible that other individual characteristics, such as training motivation, have a more important role. Incorporating these variables within the design of future studies might shed light on the training gain diff erences. Implications and limitations There are limitations to this study, which need to be addressed. Most limitations result from the convenience sampling employed. The fi rst and utmost limitation comes from a small sample size, thus, some of the insignifi cances or lack of eff ects could be attributed to a lack of statistical power. Another problem is the uneven gender distribution; thus, gender eff ects could not be controlled. Third, specifi c personality (i.e. high/low conscientiousness, high/low neuroticism) was not an “inclusion” factor; thus, the eff ects might have been stronger had we included participants with the personality profi le in the upper and lower part of the trait spectrum. Finally, although previous studies have shown eff ectiveness of this training procedure in enhancing WM in young adults (Tkalčević, 2013; Tkalčević and Vranić, 2013), adding a control group could separate the eff ect of the intervention from multiple testing eff ects. Furthermore, more often than not computerized cognitive training (CCT) in older adults suggests there is a maximal dose for CCT, after which other factors (such as fatigue) come i nto play (Lampit et al., 2014). Comparative st udies i n you nger adults, as in this study, link training schedules distributed over time with a greater training effi cacy (Penner et al., 2012). The dosage in this study and its predecessors was decided upon to closely mirror the evidenced-based effi cient training conducted by Jaeggi and colleagues ( 20 1 1 ), our procedure was mirrored to our previous control study (Tkalčević and Vranić, 2013). However, it remains likely that our results might change given a diff erent dosage or training schedule. A l t h o u g h w i t h s o m e l i m i t a t i o n s , t h i s s t u d y h o l d s a strong point in the multilevel modeling approach ( K w o k e t a l . , 2 0 0 8 ) . C o n s i d e ri n g t h e s tu d i e d c o n s tru c t s , this type of analysis has enabled the calculation of cross- level interaction eff ects and explanation of between-subject variance in growth curves by subject-level predictor ( fl uid reasoning). Also, MLM treats regression parameters from all the individual growth models (intercepts) as random eff ects for estimations thus reducing type I error in statistical inference. Future research should further consider possible moderators, which requires adequate sample sizes to allow the adequate detection of eff ects. Knowledge of such mode- rators should ultimately allo w the desi gn o f interv entions aimed at particular cognitive skills on an individual level, and with the ultimate goal to increase the effi cacy of the intervention (Buschkuehl et al., 2012). References Allerhand, M. (2016). Longitudinal data analysis. C e n t r e f o r C o g n i t i v e A g e i n g a n d C o g n i t i v e Epidemiology. http://www.ccace.ed.ac. uk/sites/default/fi les/longitudinal-2up_1.pdf Altmann, E. M., & Gray, W. D. (2008). An integrated model o f c o gni ti v e c o n tr o l in ta s k s wi t c hin g . Psychological Review, 115, 602–639. Au, J ., Sheehan, E., T sai, N., Duncan, G. J ., Buschkuehl, M., & Jaeggi, S. M. (2015). Improving fl uid intelligence w i t h tr a i n i n g o n w o r k i n g m e m o ry : A m e t a - a n a l y s i s . Psychonomic Bulletin & Review, 22, 366–377. Backman, L., Hill R.D. & Forsell, Y . (1996). The infl uence o f d e p r e s s i v e s y m p t o m a t o l o g y o n e p i s o d i c m e m o r y functioning among clinically nondepressed older adults. Journal of Abnormal Psychology, 105, 97–105. Barrouillet, P. (1996). Transitive inferences from set-inclusion relations and working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1408–1422. Brickenkamp, R. (1999). Test optere ćenja pažnje: Priru čnik za test d2 [Test of attention: Manual]. Naklada Slap. Buschkuehl, M., Jaeggi, S. M., & Jonides, J. (2012). Neuronal eff ects following working memory training. Developmental Cognitive Neuroscience, 2, S167–S179. Colquitt, J. A., LePine, J. A., & Noe, R. A. (2000). T oward an integrative theory of training motivation: A meta- analytic path analysis of 20 years of research. Journal of Applied Psychology, 85, 678–707. Conway, A. R. A., Kane, M. J., Bunting, M. F., Hambrick, D . Z., Wilhelm, O ., & Engle, R. W . ( 2005). W orking memory span tasks: A methodological review and user’s guide. Psychonomic Bulletin & Review, 12, 769–786. D a n e m a n , M . , & C a r p e n t e r , P. A . ( 1 9 8 3) . I n d i v i d u a l d i ff erences in integrating information between and within sentences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 561–584. Diamond, A. (2013). Executive functions. Annual Review of Psychology, 64, 135–168. Ecker, U. K. H., Lewandowsky , S., Oberauer, K., & Chee, A. E. H. (20 1 0). The components of working memory updating: An experimental decomposition and individual diff erences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36, 170–189. E y s e n c k , M . W . , & C a l v o , M . G . ( 1 9 9 2 ) . A n x i e t y a n d pe r for ma nce: T he processi ng effi ciency theory. Cognition and Emotion, 6, 409–434. 136 Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive performance: Attentional control theory. Emotion, 7, 336–353. Gajewski, P . D., Hanisch, E., Falkenstein, M., Thönes, S., & Wascher, E. (2018). What does the n-back task measure as w e g et o l der ? R e la ti o ns betw een w o rkin g - m em o ry m e a s u r e s a n d o t h e r c o g n i t i v e f u n c t i o n s a c r o s s t h e lifespan. Frontiers in Psychology, 9, Article 2208. Goldberg, L. R. (1999). A broad-bandwidth, public domain, personality inventory measuring the lower-level facets of several fi ve-factor models. In I. Mervielde, I. Deary, F . De Fruyt, & F. Ostendorf (Eds.), Personality Psychology in Europe, (Vol. 7, pp. 7–28). Tilburg University Press. Grimley, M., Dahraei, H., & Riding, R. J. (2008). The relationship between anxiety -stability, working memory and cognitive style. Educational studies, 34, 213–223. Hox, J. J. (2010). Quantitative methodology series: Multilevel analysis: Techniques and applications (2nd ed.). Routledge/Taylor & Francis Group. Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Shah, P. (2011). Short- and long-term benefi ts of cognitive training. PNAS Proceedings of the National Academy of Sciences of the United States of America, 108, 10081–10086. Jaeggi, S. M., Buschkuehl, M., Shah, P., & Jonides, J. (2014). The role of individual diff erences in cognitive training and transfer. Memory & Cognition, 42, 464–480. Jaeggi, S. M., Studer-Luethi, B., Buschkuehl, M., Su, Y. F ., Jonides, J., & Perrig, W . J. (20 1 0). The relationship be tw een n-b ac k perf o rman ce an d ma trix reas o nin g— implications for training and transfer. Intelligence, 38, 625–635. Kane, M. J., Conway , A. R. A., Miura, T . K., & Colfl esh, G. J. H. (2007). Working memory, attention control, and the n-back task: A question of construct validity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, 615–622. Karbach, J. & Spengler, M. (2012). Who benefi ts the most? Individual diff erences in the transfer of executive control training in younger and older adults. In A. Bröder et al. (Eds.), Abstracts of the 54th Meeting for Experimental Psychologists (TEAP), S. 64. Pabst. Kirchner, W. K. (1958). Age diff erences in short-term retention of rapidly changing information. Journal of Experimental Psychology, 55, 352–358. Komarraju, M., & Karau, S. J. (2005). The relationship between the big fi ve personality traits and academic motivation. Personality and Individual Di ff erences, 39, 557–567. Kwok, O. M., Underhill, A. T., Berry, J. W., Luo, W., Elliott, T . R., & Y oon, M. (2008). Analyzing longitudinal data with multilevel models: An example with individuals living with lower extremity intra-articular fractures. Rehabilitation Psychology, 53, 370–386. Kyllonen, P . C., & Christal, R. E. (1990). Reasoning ability i s ( l i t t l e m o r e t h a n ) w o r k i n g - m e m o r y c a p a c i t y ? ! Intelligence, 14, 389–433. Kyllonen, P. C., & Stephens, D. L. (1990). Cognitive abilities as determinants of success in acquiring logic skill. Learning and Individual Di ff erences, 2, 129–160. L a m p i t , A . , H a l l o c k , H . , & V a l e n z u e l a , M . ( 2 0 1 4 ) . Computerized cognitive training in cognitively healthy older adults: A systematic review and meta-analysis of eff ect modifi ers. PLoS Medicine, 11(11), e1001756. Logie, R. H. (2011). The functional organization and capacity limits of working memory. Current Directions in Psychological Science, 20, 240–245. Lövdén, M., Brehmer, Y., Li, S. C., & Lindenberger, U. (2012). Training-induced compensation versus magnifi cation of individual diff erences in memory performance. Frontiers in Human Neuroscience, 6, Article 141. Martocchio, J. J., & Judge, T. A. (1997). Relationship between conscientiousness and learning in employee training: Mediating infl uences of self-deception and self-effi cacy. Journal of Applied Psychology, 82, 764–773. Matysiak, O., Kroemeke, A., & Brzezicka, A. (2019). Working memory capaci ty as a predi ctor o f co gni ti v e training effi cacy in the e l der l y populati on. Frontiers in Aging Neuroscience, 11, 126. Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training eff e c t i v e ? A m e t a - a n a l y t i c r e v i e w . Developmental Psychology, 49, 270–291. Miyake, A. , F riedman, N. P . , Emerson, M. J . , Witzki, A. H., & Howerter, A. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100. Morris, N., & Jones, D. M. (1990). Memory updating in working memory: The role of the central executive. British Journal of Psychology, 81, 111–121. Naquin, S. S., & Holton, E. F. III. (2002). The eff ects of personality, aff e c t i v i t y , a n d w o r k c o m m i t m e n t o n motivation to improve work through learning. Human Resource Development Quarterly, 13, 357–376. Pauwels, L., Chalavi, S., & Swinnen, S. P. (2018). Aging and brain plasticity. Aging, 10, 1789–1790. Penner, I-K, Vogt, A., Stöcklin, M, Gschwing, L, Opwis, K., & Calabrese, P. (2012). Computerised working memory training in healthy adults: A comparison of two diff erent training schedules. Neuropsychological Rehabilitation, 22, 716–33. Redick, T. S., Shipstead, Z., Wiemers, E. A., Melby- Lervåg, M., & Hulme, C. (2015). What’s working in working memory training? An educational perspective. Educational Psychology Review, 27, 617–633. Schmiedek, F ., Lövdén, M., & Lindenberger, U. (2009). On the relation of mean reaction time and intraindividual reaction time variability. Psychology and Aging, 24, 841–857. S c h m i e d e k , F . , L ö v d é n , M . , & L i n d e n b e r g e r , U . ( 2 0 1 4 ) . Younger adults show long-term eff ects of cognitive training on broad cognitive abilities over 2 years. Developmental Psychology, 50, 2304–2310. Schwaighofer, M., Fischer, F., & Bühner, M. (2015). Does working memory training transfer? A meta-analysis including training conditions as moderators. Educational Psychologist, 50, 138–166. Soveri, A., Antfolk, J., Karlsson, L., Salo, B., & Laine, M. (2017). Working memory training revisited: A multi-level meta-analysis of n-back training studies. Psychonomic Bulletin & Review, 24, 1077–1096. A. Vrani ć, M. Martin čevi ć and V . Prpi ć 137 Predictors of training e ffi cacy during n-back training Studer-Luethi, B., Bauer, C. E., & Perrig, W. J. (2012a). Neuroticism aff ects working memory and training p e r f o r m a n c e i n r e g u l a r l y d e v e l o p e d s c h o o l c h i l d r e n . International Journal for Cross-Disciplinary Subjects in Education, 3, 640–647. Studer-Luethi, B., Jaeggi, S. M., Buschkuehl, M., & Perrig, W. J. (2012b). Infl uence of neuroticism and conscientiousness on working memory training outcome. Personality and Individual Di ff erences, 53, 44–49. Titz, C., & Karbach, J. (2014). Working memory and executive functions: Eff ects of training on academic achievement. Psychological Research, 78, 852–868. Tkalčević, B. (2013). Working memory training: Testing the e ff ects of transfer and maintenance [Master ’ s th es is ] . Faculty of Humanities and Social Sciences. T kalčev ić, B., & Vra n ić, A. (2013). Work i ng memor y t rai n i ng: T r a n s f e r a n d m a i n t e n a n c e e ff e c t s . I n G . K u t e r o v a c Jagodić, I. Erceg Jugović, & A. Huić (Eds. ), 21. days of Ramiro and Zoran Bujas: Book of abstracts (pp. 203–203). Faculty of Humanities and Social Sciences, University of Zagreb. Tziner, A., Fisher, M., Senior, T., & Weisberg, J. (2007). Eff ect s of t r ai nee cha r acter ist ics on t r ai n i ng eff ectiveness. International Journal of Selection and Assessment, 15, 167–174. Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W. (2005). An automated version of the operation span task. Behavior Research Methods, 37, 498–505. von Bastian, C. C., & Oberauer, K. (2014). Eff ects and m e c h a ni s m s o f w o r kin g m e m o ry tr a inin g : A r e v i e w . Psychological Research, 78, 803–820. Vranić, A. (2011). Test prostornog slijeda. Centar za psihodijagnostičke instrumente. Wiemers, E.A., Redick, T.S., & Morrison, A.B. (2019). The infl uence of individual diff erences in cognitive ability on working memory training gains. Journal of Cognitive Enhancement, 3, 174 –185. Zatorre, R. J., Douglas Fields, R., & Johansen-Berg, H. (2012). Plasticity in gray and white: Neuroimaging changes in brain structure during learning. Nature Neuroscience, 15, 528–536. Prispelo/Received: 31. 12. 2019 Sprejeto/Accepted: 11. 12. 2020