The general claim about novel molecularly imprinted polymers is that they are selective for their template or for another target compound. This claim is usually proved by some kind of experiment, in which a performance parameter of the imprinted polymer is shown to be better towards its template than towards interferents. A closer look at such experiments shows, however, that different experiments may differ substantially in what they tell about the same imprinted polymer’s selectivity. Following a short general discussion of selectivity concepts, the selectivity of imprinted polymers is analyzed in batch adsorption, binding assays, chromatography, solid phase extraction, sensors, membranes, and catalysts. A number of examples show the problems arising with each type of application. Suggestions for practical method design are provided.

A novel method is successfully tested for non-covalent imprinting. Conditions are used which practically exclude the formation of prepolymerization complexes. The template is cholesterol, and no so-called functional monomer is used. The polymers contain only an acrylic diester crosslinker. The porogen isopropanol prevents even hydrogen bonding between the template and the monomer in the prepolymerization solution. Despite of these apparently very disadvantageous conditions, appreciable imprinting factors for cholesterol and imprinted selectivity against some other steroids are observed, similar to other cholesterol MIPs with proven analytical usefulness.

Selectivity is extremely important in analytical chemistry but its definition is elusive despite continued efforts by professional organizations and individual scientists. This paper shows that the existing selectivity concepts for univariate analytical methods broadly fall in two classes: selectivity concepts based on measurement error and concepts based on response surfaces (the response surface being the 3D plot of the univariate signal as a function of analyte and interferent concentration, respectively). The strengths and weaknesses of the different definitions are analyzed and contradictions between them unveiled. The error based selectivity is very general and very safe but its application to a range of samples (as opposed to a single sample) requires the knowledge of some constraint about the possible sample compositions. The selectivity concepts based on the response surface are easily applied to linear response surfaces but may lead to difficulties and counterintuitive results when applied to nonlinear response surfaces. A particular advantage of this class of selectivity is that with linear response surfaces it can provide a concentration independent measure of selectivity. In contrast, the error based selectivity concept allows only yes/no type decision about selectivity.