Human beings take many times to recognize or even misrecognize some visual information because of the mismatch of colors and meanings under some situations, that is called Stroop effects. Including Stroop effects, the cognitive process of the brain is still elusive. As a first step of understanding the cognitive process, we construct and analyze a predictive coding model with hierarchical reservoir computing for modeling Stroop effects in this study. The model learns respectively colors, a shape, and a gender of normal images where a pictogram is illustrated, and that predicts their gender from the input images. Normal and abnormal images are applied to the model, whose gender predictions succeed by 90% for the normals but by 18% for the abnormals.