cognitivecontrol, task switching, dual tasks, PrP effect
in the present study, participants performed highly comparable task-switching and dual-task paradigms, and the paradigm-specific performance costs were analysed in the context of the commonly postulated core components of cognitivecontrol (i.e., working memory updating, in- hibition, and shifting). in the task-switching paradigm, we found switch costs (i.e., switch trials vs. repetition trials) and mixing costs (i.e., repetition trials in mixed-task blocks vs. single-task trials). in the dual-task paradigm, we observed a psychological refractory period (PrP) effect (i.e., task 2 [t2] performance after short stimulus-onset asynchrony [soA] vs. long soA), dual-task costs (i.e., t2 dual-task performance with a long soA in trials with a task repetition between task 1 [t1] and t2 vs. single-task performance), and switch costs in t2 (i.e., dual-task performance in trials with a switch between t1 and t2 vs. dual-task performance in trials with a repetition between t1 and t2). A with- in-subjects comparison of the performance costs showed a correlation between mixing costs and dual-task costs, possibly indicating shared underlying cognitivecontrol processes in terms of work- ing memory updating. surprisingly, there was also a correlation between switch costs and the PrP effect, presumably suggesting that cognitivecontrol, as opposed to passive queuing of response selection processes, contributes to the PrP effect.
ond was deﬁned an emotional Stroop task. Depending on the task, nonemotional (‘‘FEMALE’’ vs. ‘‘MALE’’) or emo- tional distractor words (‘‘FEAR’’ vs. ‘‘HAPPY’’) were written across the faces, generating nonemotional (e.g., when ‘‘FE- MALE’’ was written over a male face) or emotional conﬂict (e.g., ‘‘FEAR’’ written over a happy face). On a behavioral le- vel, the data of Egner et al. (2008) revealed a similar pat- tern of RT congruency effects and post-conﬂict adjustments for the emotional and the nonemotional Stroop task as for the classic color-word Stroop task, i.e. RTs were slower in incongruent than in congruent trials and the amount of the Stroop effect was smaller after pre- viously incongruent than after congruent trials. The neuro- imaging data, however, revealed a dissociation between nonemotional and emotional conﬂict processing: For non- emotional conﬂicts, the authors found that conﬂict moni- toring and cognitivecontrol was associated with dACC and DLPFC activity, respectively (see also Kerns et al., 2004 ). For emotional conﬂict processing, however, in- creased activity was found both in the amygdala and the dACC in incongruent relative to congruent trials, while en- hanced cognitivecontrol processes after conﬂict trials was correlated with activity in the rostral anterior cingulate cortex (rACC) but not in the DLPFC (see also Etkin, Egner, Peraza, Kandel, & Hirsch, 2006 ). From these ﬁndings, Egner et al. (2008) concluded that two distinct reactive control loops exist in the human brain: While the dACC serves as a conﬂict monitor both for emotional and nonemotional conﬂicts, the occurrence of post-conﬂict adjustments is associated with enhanced DLPFC activity only for nonemo- tional conﬂicts. In contrast, post-conﬂict adjustments after emotional conﬂicts are associated rather with rACC activity.
Curtis & D’Esposito, 2001; Postle, Brush, & Nick, 2004; Wolf, Walter, & Vasic, 2010). For example, Wolf and colleagues (2010) used a PI task with varying contextual demands and investigated corresponding activity in different brain regions. They proposed that increased cognitivecontrol exerted by the left dlPFC is relevant for decreasing susceptibility to PI. Secondly, the dlPFC has been associated with controlling unwanted memories and thoughts as well as with symptoms of PTSD (e.g., Anderson et al., 2004; Arnsten, Raskind, Taylor, & Connor, 2015; Clark et al., 2003). For instance, in a suppressing versus recalling task of memory contents, healthy individuals showed greatest dlPFC activation when intrusive memories needed to be controlled and individuals with negative coupling between dlPFC and hippocampus during early suppression attempts experienced fewer intrusive memories later on (Benoit, Hulbert, Huddleston, & Anderson, 2014). Thirdly, beneficial effects of anodal tDCS and detrimental effects of cathodal tDCS over the left dlPFC on cognitivecontrol have been documented in previous research (Andrews et al., 2011; Fregni et al., 2005; Wolkenstein & Plewnia, 2013; Wolkenstein et al., 2014; Zaehle, Sandmann, Thorne, Jancke, & Herrmann, 2011). Moreover, it has been shown that the effect of anodal tDCS over the left dlPFC on state rumination, as measured several minutes after stimulation termination, is mediated by an enhancement of cognitivecontrol in healthy individuals (Vanderhasselt et al., 2013). In sum, in the current study we focus on the left dlPFC as there is consistent evidence that it plays a major role in cognitivecontrol, including resistance to PI. However, it should of course be noted that resistance to PI and the control of intrusive memories and rumination can be expected to be supported by a complex inhibitory network, involving different brain regions that might also be promising targets for tDCS in future research (e.g., Anderson, Bunce, & Barbas, 2015; Badre & Wagner, 2005; Blasi et al., 2006; Caplan, McIntosh, & De Rosa, 2007; D'Esposito, Postle, Jonides, & Smith, 1999; Feredoes, Tononi, & Postle, 2006; Johnson, Saykin, Flashman, McAllister, & Sparling, 2001; Kühn, Vanderhasselt, De Raedt, & Gallinat, 2012; Nee, Jonides, & Berman, 2007).
Despite the fact that the 5-to-7 shift has been an impor- tant topic in developmental psychology research, the emerging field of developmental cognitive neuroscience has thus far been silent about it. Although functional MRI (fMRI) studies have demonstrated a close link between development of cognitivecontrol during middle child- hood and adolescence and the maturation of frontoparietal circuitry (e.g., Luna et al., 2001; Satterthwaite et al., 2013), fMRI studies focusing specifically on the age range of 5 to 7 have been sparse (but see Sheridan, Kharitonova, Martin, Chatterjee, & Gabrieli, 2014). Electroencepha- lography (EEG) studies that have looked at cognitivecontrol during this age range have revealed an age- related increase in the error positivity (Grammer, Carrasco, Gehring, & Morrison, 2014) and a positive association between the error positivity and early academic achieve- ment (Kim et al., 2016). However, nothing is known so far about the direct effect of schooling on the development of the brain network underlying cognitivecontrol.
The motivation to assume a single, central system is based on the consideration that multiple component models of WM that suggest a control mechanism that is structurally separate from WM (Baddeley & Hitch, 1974; Cowan, 1999; Engle et al., 1999; Norman & Shallice, 1986) have many drawbacks. For instance, such a system could be characterized as a homunculus (Baddeley, 1996; Miyake et al., 2000). The view of the nature of capacity limits in WM we propose here draws on single resources models that integrate an executive functioning within the framework of WM (Oberauer, 2013; Vandierendonck, 2012). Such models fractionate WM functioning into two parts. One module of WM is related to WM storage, whereas the other module involves an executive-related function. In the latter one, task set information is loaded. Since cognitivecontrol processes are goal-driven, and thus, per definition, are functionally integrated within the task set, they are supposed to call on this executive-related function. Thus, tasks always tap into the executive function when the task set involves a specific component of cognitivecontrol, such as selective attention. Important to note is that the modules for WM storage and executive function both suffer from limitations in the total amount of activation being available (Vandierendonck, 2012). If further both processes tap into the same resource, the allocation of limited resources to storage or executive function becomes crucial. In accordance with Carpenter and Just (1988), we suggest that variations in WM capacity in part reflect variations in the effectiveness with which limited resources are allocated to a given WM module, and high but not low WM capacity individuals would be efficient in doing so. Consequently, the optimal WM functioning for low WM capacity individuals might be achieved by telling individuals what process they should allocate resources to (Carpenter & Just, 1988).
Kiehl et al., 2000; Braver et al., 2001; Ullsperger and von Cramon, 2001; Garavan et al., 2002; see also Swick and Turken, 2002) tried to disentangle brain activity related to errors from brain activity related to high-conflict and were able to differentiate subdivisions of the pFMC. The focus of response conflict-related activation was found in the mesial BA 8, whereas error processing engaged a region more ventrally and anterior, in the rostral anterior cingulate cortex (rACC). Interestingly, the pre-SMA (BA 6) and adjoining dorsal ACC (dACC) areas showed activation during both processes (Ullsperger and von Cramon, 2001). Corroborating this finding of a subregional dissociation in the superior-inferior direction, Ullsperger and von Cramon (2004) reviewed main activation coordinates related to conflict and error processing from fourteen neuroimaging studies. Although response conflict seems to engage a more dorsal region of the frontomedian wall when compared to error processing, some overlap of activation was found between conditions. Thus, it remains unresolved whether these data reflect a strict functional specialization or rather a different degree of engagement of the above mentioned midline structures. The contribution of these different brain regions to cognitivecontrol and error processing is still under debate. In particular, while the conflict monitoring hypothesis proposes the view of the dACC as a conflict detector which alerts regulative processes subserved by lateral prefrontal regions, a recent study demonstrated that the dACC might be engaged in regulation processes itself (Roelofs et al., 2006). These authors observed dACC activation not only on incongruent but also on neutral trials in which no conflicting response alternatives occur.
cognitivecontrol, task switching, dual tasks, PRP eﬀect
In the present study, participants performed highly comparable task-switching and dual-task paradigms, and the paradigm-speciﬁc performance costs were analysed in the context of the commonly postulated core components of cognitivecontrol (i.e., working memory updating, in- hibition, and shifting). In the task-switching paradigm, we found switch costs (i.e., switch trials vs. repetition trials) and mixing costs (i.e., repetition trials in mixed-task blocks vs. single-task trials). In the dual-task paradigm, we observed a psychological refractory period (PRP) eﬀect (i.e., Task 2 [T2] performance after short stimulus-onset asynchrony [SOA] vs. long SOA), dual-task costs (i.e., T2 dual-task performance with a long SOA in trials with a task repetition between Task 1 [T1] and T2 vs. single-task performance), and switch costs in T2 (i.e., dual-task performance in trials with a switch between T1 and T2 vs. dual-task performance in trials with a repetition between T1 and T2). A with- in-subjects comparison of the performance costs showed a correlation between mixing costs and dual-task costs, possibly indicating shared underlying cognitivecontrol processes in terms of work- ing memory updating. Surprisingly, there was also a correlation between switch costs and the PRP eﬀect, presumably suggesting that cognitivecontrol, as opposed to passive queuing of response selection processes, contributes to the PRP eﬀect.
In the current approach, we therefore explored neuronal activity during game playing that involved simple stimulus-response mappings. We used a previous 2D game of ours [ 23 ] where participants had to catch falling objects. Those objects either fell randomly (RANDOM task), according to a known stimulus-response mapping based on color or shape (APPLY task), or according to an unknown mapping of the same type (LEARN task). We expected performance increases from RANDOM to LEARN to APPLY reflecting the different levels of task complex- ity (see Material and Methods for details). In addition, once a stimulus-response mapping is known and validated participants should perform better at catching the next objects falling. Near-infrared spectroscopy (NIRS) allowed exploring neuronal activity related to processes of cognitivecontrol during gaming. NIRS is similar to fMRI in that it reflects the hemodynamic response in cerebral vessels [ 24 – 26 ]. The observed changes in hemoglobin concentration have been recorded in different movement paradigms [ 27 , 28 ] and can even represent activations during motor imagery [ 29 , 30 ]. Compared to most fMRI scanning protocols, NIRS offers a bet- ter time resolution [ 31 , 32 ] and is more mobile and robust to movement artifacts [ 33 ]. Prior studies have shown specific hemoglobin changes in prefrontal cortices during tasks of working memory [ 34 – 37 ] and video game playing [ 38 , 39 ]. We thus hypothesized that modulating the demands of cognitivecontrol in a game task should be detectable in non-invasive NIRS signals over a network of different brain regions.
2.4 Empirical evidence for modality-specific influences in dual-tasks
Along the lines of most models discussed above, traditionally research focused on central, amodal mechanisms of cognitivecontrol. Yet, more recent research raised the issue of modality-specificity of cognitivecontrol. If central mechanisms, reflecting cognitivecontrol processes necessary to coordinate two or more tasks are generic and amodal, then the modalities involved in the individual tasks should not determine dual-task interference. A possibility to test this hypothesis is by investigating whether dual-task costs are determined by the mapping between the different stimulus and response modalities of the tasks involved. Following this logic, a series of studies was conducted demonstrating that tasks combining visual stimulation with manual responses (VM task) and auditory stimulation with vocal responses (AV task) (i.e., modality compatible tasks) could be performed in parallel after sufficient practice, whereas visual-vocal (VV) tasks combined with auditory-manual (AM) tasks (i.e., modality incompatible tasks) resulted in significant dual task costs even after extensive practice (e.g., Hazeltine & Ruthruff, 2006; Hazeltine, Ruthruff, & Remington, 2006; Ruthruff, Hazeltine, & Remington, 2006; Stelzel, Schumacher, Schubert, & D´Esposito, 2006; see Figure 1).
The aim of the current study was to investigate the effects of postural control demands on cognitivecontrol processes in concurrent auditory-manual task switching. To this end, two experiments were conducted using an auditory cued task-switching paradigm with different postural control demands (sitting vs. standing). This design allowed us to explore the effect of postural control on switch costs, mixing costs, and the between-task congruency effect. In addition, we varied the cue-based task preparation in Experiment 1 to examine whether preparation processes are independent of additional postural control demands or if the motor control processes required by the postural control demands interfere with task-specific cognitive preparation processes. The results show that we replicated the standard effects in task switching, such as switch costs, mixing costs, and congruency effects in both experiments as well as a preparation-based reduction of these costs in Experiment 1. Importantly, we demonstrated a selective effect of postural control demands in task switching in terms of an increased congruency effect when standing as compared to sitting. This finding suggests that particularly in situations that require keeping two tasks active in parallel, the postural control demands have an influence on the degree to which cognitivecontrol enforces a more serial (shielded) mode or a somewhat less selective attention mode that allows for more parallel processing of concurrently held active task rules.
As described earlier, the congruency effect is a measure for between-task interference ( Meiran, 2005 ). If participants alternate between two tasks, they must keep task sets and rules distinct enough to prevent interference. Several authors have argued for a shielding function that protects from interference and helps focusing attention on the relevant task by increasing selectivity of processing (e.g., Dreisbach and Haider, 2008, 2009 ). In incongruent trials, the irrelevant stimulus feature can activate a competing response instantiating the currently irrelevant S–R rule, which creates task interference that has to be resolved ( Kiesel et al., 2010 ). Better task set shielding should keep the congruency effect small. Conversely, less efficient task-set shielding would increase the degree of parallel processing and thus the degree of task and response competition that arises on incongruent trials. Our findings suggest that shielding is less efficient when task sets need to be switched in the mixed blocks, in particularly if cognitivecontrol processes are already occupied by a concurrent postural control task. From a slightly different perspective, one might argue that the main difference between single tasks and mixed tasks is that tasks are processed strictly serially in single task blocks and only one task set is necessary to perform the task successfully. In contrast, both task sets need to be kept active in
Luria and Meiran (2003, 2006) found T1 and T2 performance to be worse in order switch trials than in order repetition trials. Referring to this data pattern as order switch cost, they argued that the existence of this cost provides evidence for the activation of subtask order information in PRP trials and, hence, for the requirement of cognitivecontrol at a more global level than that of the subtask. Order switch costs were also found in other studies, for example, in the study by De Jong (1995) who explored preparation in overlapping-task performance and in the study by Lien and Ruthruff (2004) who investigated the effect of hierarchical task organization on task switching. Further evidence suggesting the activation of subtask order information in dual tasks comes from neuroimaging studies (e.g., Szameitat, Lepsien, von Cramon, Sterr, & Schubert, 2005; Szameitat, Schubert, Müller, & von Cramon, 2002) that report different patterns of brain activity for order repetition trials and order switch trials. In order switch trials, there was an increased activation of the lateral prefrontal cortex (i.e., cortical areas along the posterior part of the left inferior frontal sulcus and the right posterior middle frontal gyrus) that might contribute to task control.
switch costs (Karayanidis et al., 2009; Karayanidis, Provost, Brown, Paton, & Heathcote, 2011; Karayanidis, Whitson, Heathcote, & Michie, 2011; Lavric et al., 2008; Wang, Ding, & Kluger, 2015). Nonetheless, subsequent studies were able to demonstrate that the switch positivity in fact consists of several subcomponents. For example, including single-task blocks (AAAAAA) in the experimental design and comparing the associated all-repetition trials with conventional (mixed-)repetitions allows for a differentiation of the switch positivity and a slightly more centro-parietal and slightly earlier mixing positivity. This substantiates accounts proposing that indeed two separate mechanisms underlie switch-independent cue encoding and switch-specific task set reconfiguration (Goffaux et al., 2008; Jost et al., 2008; Karayanidis, Whitson, et al., 2011; Nicholson, Karayanidis, Poboka, Heathcote, & Michie, 2005). Furthermore, studies have identified other preparatory processes with distinct functional significance, such as an early frontal modulation prior to the posterior positivity that is supposed to reflect an early stage of initiation of the subsequent task-set reconfiguration (Astle et al., 2008a; Brass, Ullsperger, Knoesche, von Cramon, & Phillips, 2005; Rushworth, Passingham, & Nobre, 2005). Additionally, a stimulus-preceding negativity is reported to represent inhibitory processes that suppress the conflicting response set (Astle et al., 2008a; Astle, Jackson, & Swainson, 2008b; Brunia, van Boxtel, & Böcker, 2012; Karayanidis, Provost, et al., 2011; van Boxtel & Böcker, 2004; Walter et al., 1964). Taken together, a complex picture of preparatory processes emerges that encompasses a multitude of distinguishable control processes in task preparation.
Separately, there are concepts of self‐optimization developed by the Cluster of Excellence “Integrative Production Technology for High‐Wage Countries” at RWTH Aachen University. This thesis focuses on the human‐centered design of self‐optimizing production systems. Thereby, the main objective is related to the need for an integrated view on production systems. The developed concept also refers to the optimization of conflict resolution without focusing too much on a single element. The design of self‐optimizing systems can adapt to the objective based on the situation. Self‐optimizing means that the system will work based on simulated cognition, which signifies that the technical system is able to perform planning (semi‐)autonomously and learn from past experiences to a certain degree (Mayer et al., 2008). These systems are designed based on a continuous decision cycle: analyzing the current situation, deriving possible new system objectives, tasks and procedures and adopting the system behavior autonomously (Brecher, 2012). Thus, self‐optimizing systems require a flexible and mutable automation. The automation should be able to manage complex processes without the necessity of manual intervention based on simulated cognition. The next generation of self‐optimizing systems is able to learn and adapt their behavior. Regarding joint cognitive systems, the human operator must be considered as part of the production system. Since human behavior is much more unpredictable than a machine, the mutability of the system also has to cope with that challenge (Faber et al., 2013b).
143 einem multiplen Modell der Form ARIMA(2,1,1)×(2,1,0)24 mit den Koeffizien- ten φ1 = 0,46; φ2 = −0,1; θ1 = 0,76; Φ1 = 0,09; Φ2 = 0,8 erreicht werden. Für die untersuchten Belegungsdaten wurde mit diesem Modell ein Prognosege- winn von 16 Prozent für 3 Prognoseschritte ermittelt. Für die Anwendung in einem Multiband Cognitive Radio-System bedeutet dieses Ergebnis, dass eine Reduktion der zeitlichen Auflösung bei der Beobachtung eines Teilbandes möglich ist. In den dadurch verfügbaren Messzeiten können weitere Teilbänder beobachtet werden. Die Ergebnisse der Kapitel 3 und 4 führten auf die abschließend diskutierte Frage nach einer geeigneten Detektion und Signalisierung der Messergebnisse in einem Multiband Cognitive Radio-System. Dazu wurde in Kapitel 5 ein Overlay-Szenario vorgestellt, in dem GSM1800 als Lizenznutzersystem angenommen wurde. Das an- genommene Mietnutzersystem basiert auf dem Standard IEEE 802.16, der auch als WiMAX bezeichnet wird. Aufgrund der durchgeführten Adaptionen sowohl der physikalischen Schicht als auch der Rahmenstruktur wird das betrachtete Mietnut- zersystem aWiMAX-System bezeichnet. Anhand dieses Beispiels wurde gezeigt, dass ein Multiband Cognitive Radio-System im Frequenzbereich des GSM1800 Downlinks eingesetzt werden kann. Der vorgeschlagenen Rahmenaufbau ermög- licht sowohl die Signalisierung der Lizenznutzerdetektion innerhalb des aktiven Frequenzbandes als auch die Übermittlung der mittleren Auslastung benachbarter Teilbänder. Dazu wurde das aus der Literatur bekannte Boosting-Verfahren für den Multiband-Einsatz erweitert. Die Forderung nach einer flexiblen Anpassung der zeitlichen Auflösung bei der Beobachtung wurde durch den Einsatz unterschiedli- cher Detektionsstrategien erfüllt. Dabei wird auf die verteilte Detektion innerhalb des Multiband Cognitive Radio-Systems und die Prognose der Belegungswerte zu- rückgegriffen.
Aspects of cognitive functioning, such as memory, attentional performance, or face recognition, are impaired in patients with schizophrenia, and can lead to significant disabilities in occupational, social, and economic functioning (see Keefe and Harvey, 2012 , for a review). These particular impairments often persist throughout all stages of illness ( Censits et al., 1997 ). Antipsychotic agents often improve positive symptoms, but treatments that improve social and cognitive functioning are still warranted ( Keefe et al., 1999 ). Cognitive skills have become an important therapeutic target in patients with schizophrenia, and are associated with long-term outcomes and prognosis ( Purdon et al., 2000 ). Psychological strategies, such as cognitive remediation therapy, have been introduced to improve cognitive deficits ( Turkington et al., 2004 ) and should target different symptom domains, e.g., executive function, attentional performance, and aspects of memory ( Kurtz et al., 2001 ). Such improvements of specific cognitive functions ( Medalia and Lim, 2004 ) can be expected to normalize dysregulated neural activity ( Kurtz, 2012 ; Penadés et al., 2013 ). Although one may expect that addressing a supposed neural dysregulation in a direct manner should lead to improvements in cognitive functions, so far, none of the existing therapeutic approaches directly addresses the underlying neural deficits ( Turkington et al., 2004 ).
The systematic and rigorous assessment of key cognitive competencies is of high scientific and societal relevance, as is the availability of high-quality data on cognitive competencies. It is now widely accepted that in modern knowledge societies the economic prosperity of individuals, communities, and countries is associated with the cognitive competencies acquired. In order to make well-informed decisions, politicians and educational authorities need high-quality data about the effectiveness of formal and non-formal educational environments. Similarly, researchers need strong data to test complex theoretical models about how individual biographies are shaped by the interplay between individual and institutional affordances and constraints. Some questions that require high-quality data on cognitive competencies include: Has high school students’ mathematics and reading literacy generally increased or decreased in recent years and decades (see Becker et al. 2006)? Are
subsystems that are triggered by the evaluation of an external stimulus as being highly relevant to major goals and concerns of the individual. During this crucial evaluation process, that is controlled by the cognitive component, the stimulus is appraised on several different dimensions that Scherer (2001) calls stimulus evaluation checks (hereafter, these checks will be referred to as appraisal dimensions). His proposed 16 appraisal dimensions are further subdivided into four classes of information that determine how relevant an event is for the individual (relevance detection), which consequences an event has and how these will affect the individual (implication assessment), how well the individual can cope with potential consequences (coping potential determination), and how important the event is in regards to the individual self-concept and social norms (normative significance). The outcomes of these dimensions are believed to be highly subjective and to be depending exclusively on the individuals’ personal perception of the stimulus. In contrast to some other appraisal theories (e.g., Lazarus, 1991) that assume the outcome of appraisals to be partly categorical, Scherer (2001) postulates that appraisals are evaluated on a continuous scale with a potentially infinite value range. He further claims that the appraisal process is iterative and that the 16 dimensions are appraised in a specific order. See Table 1 for a short description of all appraisal dimensions in their assumed order of occurrence. The proposed sequentiality of appraisals, which is unique to Scherer's (2001) appraisal theory, is thought to ensure the economy of the cognitive appraisal process. He assumes that all appraisals incorporated in relevance detection, such as suddenness, pleasantness, and goal/need importance, are rather low-level and hence fast mechanisms that fall back on attention, memory, as well as motivational processes. Appraisals appearing later on in the process are thought to be more complex cognitive evaluations that are consequently costlier and require functions like reasoning and the evaluation of self-image. The first appraisals determining the relevance of a stimulus to the individual, therefore, act as a filter that decides whether further expensive processing of the stimulus is needed. Only when a certain threshold is surpassed with these appraisals, additional processing through other appraisals is initiated.
only applicable to this small share of the Bangladeshi labor market, our paper still contributes to a better understanding of which skills are required in the part of the labor market likely to be the primary driver of faster GDP growth and poverty reduction.
We find that non-cognitive skills do not affect the selection of the hiring channel, but they do have a direct effect on wages, after correcting for initial selection. These significant correlations are not visible in simple OLS regressions and differ by hiring channel; they therefore illustrate the benefit of first taking into account selection into different hiring channels. For those hired formally, the wage returns to an additional standard deviation of openness to experience is 2.5 percent. However, not all non-cognitive skills benefit from positive wage returns: those hired formally are punished for having higher levels of conscientiousness and hostile attribution bias. Looking at differential results by occupation shows that the result for conscientiousness is driven by white collar workers, among whom other traits, such as creativity or flexibility are arguably more important than diligently completing tasks. Among those hired through networks, hostile attribution bias also shows a negative wage return. Network hires can benefit from a positive return to emotional stability though with a one standard deviation increase in the trait associated with a 1.9 percent wage increase. These are considerable returns, especially when compared to average returns to education in developing countries. Returns to skills thus seem to align with the task content of work.