Learning to read is one of the most important cognitive milestones in the human social environment. One of the most accepted models explaining such process is the Double-Route Cascaded Model. It suggests the existence of two reading strategies: lexical and sublexical. In the Spanish language there are some contradictions about how these strategies are applied for reading. In addition, there are only a few studies dealing with the analysis of shifts between them, achieving a fluent reading process. In this paper we use a reading task including words and pseudowords for characterizing the cost of shifting between reading strategies in children with developmental dyslexia and normal controls. Our results suggest the presence of both strategies in these two experimental groups. In controls, both strategies become more efficient in correspondence to the increased exposition to written material. However, in children with developmental dyslexia only the lexical strategy exhibits such improvement. Their also point to a low cost for shifting between strategies in controls and a much more significant one in children with developmental dyslexia, differentiating subgroups with distinct shifting patterns.