 The backbone extraction technique was applied to a network of phonological word forms from English. This technique reduced the number of edges by 68.5%, resulting in a subnetwork of important words. These words were found to be more frequent in the language, shorter in length, similar to more phonological neighbors, and closer to other words than those that did not survive the backbone extraction procedure. It was suggested that these words could form a kernel lexicon, providing guidance for clinicians and researchers on which words to focus on to facilitate normal development or to speed up rehabilitation efforts. Additionally, this technique may prove useful in other applications of network science to the speech, language, hearing and cognitive sciences. This article was authored by Michael S. Veidvich and Mary Sayle.