Generally, picture books targeted at young children rely on single picture presentations. A picture illustrates an object or an action or a property. Then, a word is associated with this picture. This associative mode of teaching novel words considers word learning as a mechanism in which the word underlying a concept and the concept are connected via simple associations. This is far from current research tells us: children produce numerous extension errors (e.g., under- or over-generalizations). There is now ample evidence that the opportunity to compare several exemplars to the same target category name (e.g., several apples rather than one apple) gives better results in terms of word extension (e.g., Thibaut & Witt, 2015). Comparisons promote extensions that are based on deep semantic commonalities rather than on superficial features. The central idea is that comparisons are invitations to align objects on many properties, starting on easily accessible ones and, later, with deeper conceptually based regularities. In this paper, we will review the evidence regarding the role of comparisons in novel word learning. We will also review several conceptions of comparison effects such as progressive alignment or concreteness fading. We will illustrate situations these different comparison modes and suggest how picture books devoted to word learning might be implemented in e-learning.