Cognitive remediation covers a broad range of interventions that are heterogeneous with respect to a number of parameters. Importantly, there is considerable variation in the cognitive functions targeted in training programs. The dominant research focus in the 1980 s and 1990 s was on training procedures addressing a particular construct or even a specific task. Most prominently this included sustained attention based on findings in the Continuous Performance Test and executive function based on Wisconsin Card Sorting Test performance [9,10]. These studies were mostly focused on the question whether cognitive deficits can be remediated through training. Recently, more comprehensive training packages addressing a set of target functions have dominated the literature (e.g. [5,11]). This goes along with a shift in outcome measures. After many of the earlier studies sought to demonstrate improvement on the task trained, a broader effect on neuropsychological test performance has subsequently been considered a condition for improvement of patient relevant outcomes [12]. There is also a growing consensus that trials aimed at improving cognition should assess functional outcome directly or through an appropriate proxy measure [13]. Accordingly, functional outcome measures have been included in most recent trials of cognitive remediation (e.g. [14,15]).

Despite this rapidly developing body of research it is still a matter of discussion, which cognitive functions should be emphasized for successful cognitive remediation [16]. Interestingly, the earliest studies of cognitive remediation in schizophrenia have addressed this question to some extent. Wagner trained patients on a stimulus discrimination task with and without requirement for abstraction, but did not find a consistent advantage of one form of training [17]. Bellack and colleagues compared trained participants on either the Wisconsin Card Sorting Test or the Halstead Category Test. The authors could show that both groups improved on the non-trained test. However, these tasks involve strongly related cognitive operations and a differential effect on other cognitive functions was not the goal of the study [18]. Another line of research focused on strategies taught during training [19,20]. However, these latter studies have not included comparisons between training of different functions or tasks. Thus, it is still an open issue whether the training of certain specific functions is more effective than training of other functions. This question is pertinent in the clinical context, where therapists often employ a mix of training interventions adapted to setting and patients.

One strategy to approach this question is to relate change in specific cognitive functions to change in a functional outcome parameter. Recent studies have suggested that change in executive functions may best predict improvement in social or daily functioning and should thus receive emphasis in cognitive remediation [21,22]. Planning and problem-solving have received increased interest, because recent developments in the assessment of executive functions with high ecological validity have been applied to the study of patients with schizophrenia [23,24]. Interestingly, planning performance on tasks with real-world approximating interface and complexity has been associated with functional outcome and related proxy measures [25-27]. This includes overall performance on the naturalistic action test, community functioning and global assessment of functioning. These studies have suggested a particular role for planning and problem-solving in cognitive remediation, but have so far not provided direct evidence.

A more direct strategy to define target cognitive functions would employ head-to-head comparisons between training of specific cognitive functions. This approach could provide direct evidence for emphasizing specific cognitive functions over others. However, this type of comparison has only been conducted by Medalia and colleagues, who compared problem-solving training with memory training and treatment as usual in a sample of hospitalized patients with chronic schizophrenia [28]. Participants in the problem-solving remediation group worked under individual supervision with the software program Where in the USA is Carmen Sandiego? This educational software was selected, because it requires a range of problem-solving skills and was considered to promote intrinsic motivation. Patients who received ten sessions of problem-solving training showed greater improvement on problem-solving skills required for independent living. In contrast, patients receiving memory training did improve on the trained tasks, but not in functional outcome or executive functions [29]. Thus, this study provides direct evidence for a differential effect of different targets for intervention. However, the authors note important issues to be addressed in further research. First, it is an open question whether these findings in chronic inpatients can be generalized to less impaired patient groups. Second, final sample size was limited to less than twenty Participants in each treatment group. Third, it is an open issue whether their results pertain to the specific intervention or can be generalized to training of problem-solving in a broader sense.

Thus, the two approaches for defining the focus for cognitive remediation suggest executive functioning and more specifically planning and problem-solving as treatment targets. Planning and problem-solving can be conceived as higher executive functions, which require the integration of basic cognitive functions [30]. A crucial question is whether training of these higher-order functions strongly requiring integration provides an additional benefit over a training restricted to the basic cognitive functions (e.g. processing speed, attention, memory, lower-level executive functions). More generally, the present study addresses the question which level of cognitive functioning should be targeted.

In order to train patients on planning and problem-solving, we used the software package Plan-a-Day, which is based on an earlier concept by Funke und Krüger that has been adapted for psychiatric and neurologic patients [31]. In brief, participants are given a set of errands for one day that are described by location, time, action and importance. Participants have to interactively construct a plan for this set of errands, taking priorities and timing conflicts into account. The training can be delivered in individual and group format. In the present study, small groups of no more than five patients worked with the therapist. The comparison group trained on the basic cognitive functions processing speed, memory and attention/concentration, which have all been consistently shown to be impaired in patients with schizophrenia [32,33]. These training tasks were carefully selected to not include planning and problem-solving components.

The aim of the study was to compare the effectiveness of two different approaches to cognitive remediation: targeting planning and problem-solving versus basic cognition. To our knowledge, this type of head-to-head comparison has not been performed for cognitive remediation in a rehabilitation setting. All patients received training parallel to a three-week inpatient work therapy, which was similar for all patients. We used a measure of functional capacity as a proxy measure for functional outcome. Functional capacity is assessed under standardized conditions and has been shown to be the most consistent predictor of functional outcome [2].

The study addressed two related research hypotheses:

(1) Planning and problem-solving training leads to stronger improvement of planning ability than training of basic cognition.

Both groups improved on measures of cognitive functioning and functional capacity. We observed improvement in both patient groups in the learning ability subscale and total score of the O-AFP. The changes in O-AFP scores were above the cut-off, indicating reliable change. These findings are consistent with previous studies showing beneficial effects of programs including cognitive remediation and broader rehabilitation measures [5,7,8]. However, the interpretation of these findings is limited by the lack of a control group not receiving any cognitive intervention. Therefore, it is not clear whether our training interventions constitute a causal factor in these general improvements. The first alternative explanation to be considered is unspecific treatment effects resulting for example from hospitalization and medication. However, patients in the study were clinically stable and normally do not present short term fluctuations in performance. Another important issue is a possible effect of the intensive work therapy program on functional capacity as well as cognitive functioning. Beneficial effects of rehabilitation programs including work therapy on the OAF-P functional capacity measure have been demonstrated, although over a longer time frame [53]. Furthermore, Bell and colleagues have suggested that work therapy alone can improve cognitive functioning as it challenges memory and other cognitive functions [54]. However, to our knowledge no study has compared work therapy with a control condition in its effect on cognition.

The study's main focus was a differential effect of the training interventions on cognition and functional capacity. Regarding cognitive performance, the planning and problem-solving training lead to stronger improvement on Plan-a-Day solution time. This finding suggests that the intervention was effective at improving planning abilities. A critical objection could attribute this effect to the training of a similar task in the remediation program. However, the Plan-a-Day diagnostic and training versions differed considerably on a number of characteristics such as user interface and problem types. Therefore, although this effect might partially result from similarities between tasks, it may indicate some improvement on planning and problem-solving. There were no differential effects on the other planning tests, which address this construct on a less complex level. This difference in complexity might explain the difference in effects. In the training program, participants learn to deal with planning demands typical for real-world environments, for example involving goal conflicts requiring to skip one element. These are strategies, which are unlikely to be helpful in tasks like the Tower of London, which always have a complete and unequivocal solution.

In addition, we found a significant main effect of time for Plan-a-Day and Planungstest, suggesting that participants in both groups improved in planning ability. A critical objection would attribute this finding to a task repetition effect, although different versions of the tests were employed at both measurement points [55]. Alternatively, both the training of a more complex planning task and a set of less complex basic cognition tasks might lead to a similar improvement through different mechanisms. Overall, our results suggest that some deficits in planning and problem-solving of patients suffering from schizophrenia can be improved by a cognitive training program within three weeks. The advantage of a specific training of these functions was limited to the outcome measure most closely related to the training program. However, the improvement of the planning and problem-solving group specifically on the task most closely approaching real-world requirements suggests a potential for successful generalization to functional outcomes.

In an exploratory analysis, we addressed the issue of change in basic cognitive functions. A significant time × group interaction was only observed for reaction time in the neutral condition of the Stroop task, suggesting an advantage for basic cognition training. This result has to be viewed with caution, because we did not correct for multiple comparisons due to the exploratory character of this analysis. Reaction time in the neutral condition is a relatively pure measure of processing speed, which was also trained in the basic cognition training group. This suggests some degree of generalization across measures of processing speed, but not to other cognitive measures.

An important finding of the study is the absence of a significant differential effect of the two training programs on functional capacity. This result was observed despite the fact that the planning and problem-solving group had more contact with the trainer and explicitly practiced transfer to daily activities. Although there is meta-analytic evidence for an effect of cognitive remediation on functional outcome or respective proxy measures, this issue still remains controversial in the light of well-conducted studies with negative results [4,15]. Thus, one way to explain the absence of a differential effect would be that none of the two interventions had an effect on functional capacity.

However, Medalia et al. observed significant improvements on the Independent Living Scale specifically for the problem-solving intervention [28]. It has to be noted that our sample size was about twice as large in each treatment group and should have resulted in greater power to detect significant differences. Therefore, other differences between the studies need to be considered to explain the discrepant findings. First of all, it is important to consider similarities and differences between our intervention and the one employed by Medalia and colleagues. While both studies addressed problem-solving, our study explicitly focused on planning as a key cognitive function. In the Medalia study, planning was clearly involved in the problem-solving intervention, but a broader set of cognitive functions was likely required, although not explicitly specified. An important issue in the classification of cognitive remediation techniques is the amount of strategy teaching involved [56]. In both studies, participants in the problem-solving group were actively supported in the use of efficient problem-solving strategies. In contrast, strategies for compensating existing cognitive deficits were not explicitly trained in either study. Thus, both problem-solving interventions fill the middle ground on a continuum from drill-and-practice to compensatory approaches. Lastly, Medalia and colleagues place a strong emphasis on promoting intrinsic motivation through an engaging task environment and personal feedback. Although this was not the major theoretical background for the development of Plan-a-Day, similar elements can be found in our training task. However, in our study patients trained in small groups instead of individual training, which might have led to less individualized support and feedback. Task motivation did not differ between the two interventions, which in turn might have contributed to the observed lack of differences.

In addition, a number of factors relating to the setting and the intervention have to be considered. First, in contrast to the chronic inpatient sample in the Medalia et al. study, we included patients who were living in the community before elective admission for a treatment program promoting return to work. In addition, most patients had a relatively short duration of illness with mild impairment in cognitive functioning. A tentative interpretation of both studies would suggest that more severely impaired patients benefit more from problem-solving training in comparison to other trainings, while higher-functioning patients do not show this differential effect. Second, the duration and overall exposure to the intervention might have been too limited to produce differences between treatment groups on functional capacity. Our study was shorter than most studies of cognitive remediation (e.g. [5,6,57]), but the overall treatment exposure was larger than in the problem-solving study by Medalia et al. Nevertheless, the transfer to functional capacity in a work therapy setting might require a longer time frame. Second, in contrast to the study by Medalia, our patients participated in a broader rehabilitation program including intensive work therapy. In this enriched environment, the specific effect of a differential cognitive intervention might be more difficult to detect. Bell and colleagues have suggested that under these circumstances, a differential effect might only emerge after other treatments and supports are withdrawn [54]. Third, the control conditions differed between the two studies. In our study, the control group trained on a set of three different functions, which might have increased the effects of the basic cognition training. This combination of training targets is now implemented in most remediation programs and might be advantageous for generalization to functional outcome.