'Domain-specific is wider than specific domain knowledge' Kalyuga (2013)

 'Developing learner ability to apply knowledge in relatively new situations is an important aim of learning and instruction, and many instructional programs and materials explicitly state this goal. However, not many of them demonstrate a consistency in achieving this goal using sufficiently well-specified instructional procedures and techniques. According to a popular view, transferable knowledge and skills in complex domains result from problem solving experiences (e.g. Inagaki and Miyake 2007). At the same time, there is evidence indicating the importance and effectiveness of explicit learning generalized theoretical frameworks and abstract conceptual knowledge for deep understanding of tasks and en hancing transfer capabilities (Karpov and Bransford 1995; Hinds et al. 2001). Gick and Holyoak (1983) convincingly demonstrated the value of the acquisition of abstract schémas as mediators of analogical transfer between very different task areas.'

'Many studies in expert problem solving have demonstrated that relying on highly contextualized domain-specific knowledge associated with concrete situations is the most effective way to solve problems in familiar task areas and that expert performance is largely based on the acquisition of organized domain-specific knowledge structures (schémas; Chi et al. 1982). ... For example, the game of chess has traditionally represented a classical example of expertise based on the acquisition of a huge amount of highly specific knowledge structures (chunks) corresponding to various concrete game situations (De Groot 1965; Chase and Simon 1973). However, more recent studies have indicated that experts use not only concrete chunks but also more general patterns, such as "templates" (Gobet and Simon 1998) or generalized chunks (Walczak and Fishwick 1997) representing typical classes of chess positions with substantial variations in concrete board locations. Based on his observations of the behaviour of rats in maze situations, Tolman (1948) was one of the first to discuss the role of broad vs. narrow cognitive constructs (cognitive maps) in both animal and human learning. 

Domain-specific knowledge that is applicable only to a limited range of tasks represents the most powerful tool for dealing with these tasks in terms of efficiency of achieving immediate results. On the other hand, knowledge that can be applied more broadly is likely to be less efficient for solving specific problems since it would require additional searching, reasoning, and elaborating activities, whilst specific knowledge directly leads to a definite solutions.' pp.1478-1479.

Kalyuga, S. (2013). Enhancing transfer by learning generalized domain knowledge structures. European Journal of Psychology of Education, 28(4), 1477–1493. http://www.jstor.org/stable/23580919