Skip to main content
. Author manuscript; available in PMC: 2023 Jul 1.
Published in final edited form as: Psychol Rev. 2021 Aug 26;129(4):640–695. doi: 10.1037/rev0000313

Figure 5.

Figure 5.

Top: early in training the model registers an ‘incorrect’ association between the red object and the word ‘bosa’ (word 1). Middle: snapshots of the model’s memory traces taken every 5 training trials show gradual learning of the correct associations. Bottom: late in training the model does not form the incorrect association as in the top panel.