Due to the expansion of cities and the increasing number of large-scale and complex public spaces, there is an increase in public signage. Moreover, the information described on these signs tends to be diverse and complicated. Complex signs that contain multiple destinations or other information must be considered to determine not only the proper size, color, etc. but also the most effective arrangement of design elements. In the previous research, the cognitive utility of complex public signs was estimated using computer simulation software. In the current research, we focused on the objective estimation of the effectiveness of the results obtained in the previous research utilizing an eye mark recording system. Two cognitive engineering experiments clarified five points for improvement in the usability of complex signs, as follows: 1) Parallel construction of characters and pictograms is more efficient. 2) Grouping elements result in rapid recognition of information chunks. 3) Visual characters and pictograms are effective, along with proper density of information. 4) Specific arrangement of sign arrows is effective. 5) Figures on signs influence the sequence of information searches.