Crosscutting concerns associated with the non-functional requirements and system constraints contribute to the growing complexity of embedded software. Techniques, such as Aspect-Oriented Software Development, propose units of modularization to encapsulate their handling. A key design issue in embedded software is related to the process of locating, identifying, and specifying the Crosscutting Concerns Occurrence Points (CCOP). This paper reports on an empirical evaluation that compares four notations, namely JPDD, Theme/UML, AspectJ, and Tracematch to specify the selection of CCOPs. A set of metrics to quantify the perceptual properties of these specifications was proposed based on the Physics of Notation (PoN) conceptual framework. Three embedded automation systems have been specified using the mentioned notations and evaluated according to the proposed metrics set. Experimental results show how the used notation impacts the comprehension of the specification, whether in graphical or textual form. The results show poor discriminability in the cognitive effectiveness of visual representations. Hence, using visual notations to specify CCOP selection seems not to facilitate the understanding as one would expect.