Table 8.

Pattern of results that the model must be able to account for and proposed mechanisms

Recall
Experimental result	Mechanism
Low frequency words and high frequency words are equally likely to be recalled.	Links from episodic context to lexical or semantic representations of words are independent of frequency
Low frequency phrases and high frequency phrases are under some conditions, equally likely to be recalled at least in part	Links from episodic context to lexical or semantic representations of words (and therefore phrases) are independent of frequency
High frequency phrases are more likely than low frequency phrases to be completed once one word has been recalled	Associations between the words within the lexical-semantic system are stronger in high frequency phrases
Concrete phrases are easier to recall than abstract phrases	Concrete phrases have more active features, so the associations between a new episode and a concrete phrase is stronger
Recognition (Jacobs et al., 2016)
Experimental result	Mechanism
Low frequency words are better discriminated than high frequency words	Studied high frequency words suffer from more interference from prior episodes
High frequency phrases get more “yes” responses regardless of whether they were studied or not (a bias)	Associations between the words within the lexical-semantic system are stronger in high frequency phrases, contributing to greater familiarity
High and low frequency phrases are equally well discriminated	There are many more episodes sharing a word in a phrase than the whole phrase. Thus, interference from other phrase episodes is minimal.
Low frequency words facilitate phrase discrimination	Compositional phrases access word episodes, so high frequency words within phrases generate more interference just as they do in recognition for single words
Concrete phrases are better discriminated than abstract phrases	Concrete phrases have more active features, so the associations between a new episode and a concrete phrase are stronger