You need to know: There is a causal relationship between structural knowledge and control performance in complex problem solving tasks Natassia Goode a,b, ⁎, Jens F. Beckmann b a University of Sydney, Australia b Accelerated Learning Laboratory, University of New South Wales, Australia article info abstract Article history: Received 17 July 2009 Received in revised form 17 December 2009 Accepted 15 January 2010 Available online 24 February 2010 This study investigates the relationships between structural knowledge, control performance and fluid intelligence in a complex problem solving (CPS) task. 75 participants received either complete, partial or no information regarding the underlying structure of a complex problem solving task, and controlled the task to reach specific goals. Control was best when complete structural information was available and was not better than random when no information was provided. In comparison to previous studies, a moderate to strong correlation between fluid intelligence and control performance was observed across all conditions. It appears that effective complex problem solving requires a combination of task-specific knowledge and abstract thinking skills. Crown Copyright © 2010 Published by Elsevier Inc. All rights reserved. Keywords: Complex problem solving Dynamic systems Knowledge acquisition Control performance Fluid intelligence The production of goods in a factory, managing an economy, and driving a car, could all be described as complex, dynamic systems of interdependent relationships between variables. Presumably, once the causal relationships between variables have been discovered, this information could be used to control the outcomes of the system, or change the system, although as yet this has not been established empirically. There might also be individual differences in how well problem solvers are able to understand or utilise such information. The aim of this study is to determine whether the amount of information problem solvers have about a system has a causal impact on how well they can control it, and whether their level of fluid intelligence co-determines the extent to which this information can be applied. Several computer-based complex problem solving (CPS) tasks, sometimes referred to as simulations or micro-worlds, have been developed to represent the key features of dynamic systems (e.g. Dörner, 1987; Funke, 1992). They consist of a number of input and output variables that are represented in a computer program. The values of inputs can be changed, which affect the values of the output variables through a set of mathematical equations. These tasks are “dynamic” because the values of the outputs change in response to user input and independently over time. In order to study the role of structural knowledge in system control, often, problem solvers are first required to determine how the inputs affect the outputs (i.e. exploration phase). This is referred to as the underlying structure of the problem. They then try to control the system by manipulating the input variables to reach and to maintain defined goal states of the output variables (i.e. control phase). In complex problem solving research a common assumption is that problem solvers' control of a task is dependent upon their knowledge of the underlying structure of the problem (Funke, 2001). Indeed, the available correlational evidence supports this assumption. Funke and Müller (1988) found that control performance and knowledge of the underlying structure of the task were significantly positively correlated (r = .41), as did Beckmann and Guthke (1995) (r = .51), Vollmeyer, Burns, and Holyoak (1996) (r = .57 and r = .65), Kröner, Plass, and Leutner (2005) (r = .77 and r = .61) and Kluge (2008) (r = .82). Intelligence 38 (2010) 345–352 ⁎ Corresponding author. School of Psychology, University of Sydney, Sydney, NSW 2006, Australia. Tel.: +61 423 184720; fax: +61 2 9931 9199. E-mail address: ngoode@psych.usyd.edu.au (N. Goode). 0160-2896/$ – see front matter. Crown Copyright © 2010 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.intell.2010.01.001 Contents lists available at ScienceDirect Intelligence