Description: The research results presented refer to the evaluation of knowledge structures after a teaching-learning process. A computer software designed by the authors allows the visualization of complex knowledge structures obtained from responses to a standard multiple choice exam properly designed to encompass the knowledge structure information. The analysis of the information provided by this kind of mapping permits a fine account of: (a) number of correct concepts, (b) links between concepts in the knowledge structure, and (c) number of non correct concepts. These three parameters, presented together visually, provide an evaluation tool very efficient on determining the erroneous elements in the information structure achieved after the learning process. Such information is quite useful to correct the learning errors (and consequently improve the learning result) and also to evaluate, in a precise manner, whether the learning goals are achieved or which are the aspects to be improved to reach those goals. Quantification alternatives of the learning-moment (proximity of the information structure to the optimal one expected) are provided too. The underlying theoretical framework (see references) is the knowledge mapping theory and, more concretely, the work of the Swedish researcher Peter Gärdenfors on conceptual spaces. The central idea is that knowledge cannot be represented as a list of spared items; instead, the representation as a whole (a structure) provides the cues that lead to meaning (significant learning) and allows the prognosis of future acquisitions. Methodology: A sample of 148 university students responses to a multiple choice exam has been analyzed, obtaining the distribution of information structures parameters (see previous paragraphs). A computer software, developed for that purpose, combines the responses in a coherent knowledge structure and obtains further information (meaning nuclei, conceptual clusters, and global relations). The distance between current knowledge structures and expected (optimal) structures are computed and conceptual errors systematically accounted. This information is also contrasted with traditional assessment (correct vs. non-correct answers) results. Conclusions: Results show that effective learning is rather under-estimated by traditional evaluation methods. Individual knowledge structures variation is far larger than assessment scores, and weaknesses in individual learning differ a lot. Even thought a compound learning structure (that accounts for all the individual structures) can be computed, showing the general effectiveness of the teaching activity, a strong interaction emerges between each student cognitive traits (former knowledge, processing resources and cognitive style) and teaching effects.