Analyzing larger sample sets with rapid methods: Incomplete-block designs with free-sorting and free-linking tasks


TR Number



Journal Title

Journal ISSN

Volume Title




As rapid, holistic methods for similarity and description—such as sorting and projective mapping—have grown in popularity, a limiting factor is the number of samples that can be presented to subjects: more than 25 food samples decreases the quality and stability of results. While incomplete-block designs could address this, their use has not been developed for these holistic methods. In this paper we present an empirical investigation into the use of incomplete-block designs with free sorting and the newer free linking. We compare these two methods because while their results are comparable, the cognitive tasks are different, and thus their suitability for incomplete-block designs may differ. We evaluated the effects of incomplete-block designs in two studies. In Study 1, 20 subjects evaluated 6/10 chocolate bars by free linking in an incomplete-block design, with each subject completing 2 blocks; results were compared to a complete-block evaluation of the 10 bars by free sorting and free linking. In Study 2, a total of 90 subjects evaluated 62 terms from a chocolate flavor-wheel in 3 conditions (between subjects): free sorting with complete blocks (N = 30, all 62 terms) and free sorting (N = 30) or free linking (N = 30) with 3 incomplete blocks of 16/62 terms. We introduce a novel method to evaluate stability for the incomplete-block designs that we call “pairwise simulation.” From Study 1, we find that pairwise simulation provides adequate stability estimates and that, with sufficient pairwise cooccurrences, free linking with incomplete blocks produces results that are comparable to free sorting or linking with complete blocks. From Study 2, we demonstrate that free linking with incomplete blocks can produce high quality results from a large sample set, maintaining the increased discrimination capacity that marks free linking in general, and that with incomplete blocks, free linking is likely to be more stable than free sorting. This research demonstrates that incomplete-block designs can be used with free linking, and also provides a new, effective method through pairwise simulation for evaluating stability with incomplete-block designs, which cannot be resampled using standard bootstrapping approaches.



Free sorting, Free linking, Incomplete-block designs, Rapid methods, Sensometrics, Simulation methods