Similarity is known to affect memory. Visual item recognition refers to tasks where participants study a set of visual stimuli, and have to determine whether a probe item matches one of the items in the study set. This task is naturally sensitive to similarity effects and can well be described using "summed similarity models." These models posit that a participant will endorse a probe item as a target (member of the study set) when the sum of the similarities between each study item and the probe exceeds a decision threshold. Although the use of summed similarity models is well established, they have not yet been related to neural data.

In this dissertation, I will focus on the relation between brain oscillations recorded through human electroencephalography (EEG), visual item recognition, and summed similarity models. I will start by discussing the similarities and differences between various methods to study brain oscillations.

Then I will show that summed similarity correlates with low-frequency theta (4–9 Hz) activity recorded using scalp EEG. Conversely, proactive interference, an increase in task difficulty due to recent presentations of the same study items, is related to the amplitude of 28–90 Hz gamma oscillations. Scalp EEG suffers from low spatial resolution. I obtained intracranial EEG from epilepsy patients suffering from pharmacologically intractable epilepsy, who had electrodes implanted for clinical purposes. I will show that intracranially recorded oscillations also show correlations with summed similarity, and differentiate between probe-item similarity and item-item similarity in the medial temporal lobe. I will also show that with memory load, high-frequency oscillatory power in the medial temporal lobe increases and low-frequency oscillatory power in parietal and perceptual areas decreases.

Together, these findings show first, that similarity plays an important role in human memory. Second, summed similarity models can be directly related to oscillatory brain activity, particularly in the theta band. This opens up directions for future research, relating mathematical models of cognition to neural activity.