Our paper on discrimination difficulty during perceptual learning was just published in Psychonomic Bulletin & Review! Really short, entirely exploratory paper describing a surprising finding of a *reversal* of the target-to-distractor similarity effect with training. Thread.

Popov, V. & Reder, L. (2020). Greater discrimination difficulty during perceptual learning leads to stronger and more distinct representations. Psychonomic Bulletin & Review

Publisher link: https://bit.ly/36ELbee 
PDF: https://bit.ly/2AeR3yD 
Data & Code: https://bit.ly/3eyJkdP 

Intuitively, the greater the similarity between two different objects, the more difficult it should be to distinguish them. Research using the visual search task has repeatedly confirmed this intuition - when a target is more similar to distractors in the search array... 1/

... accuracy decreases, response times increase, and more errant saccades are made to the highly similar distractors. This negative target-to-distractor (TD) similarity effect appears to be robust and has played a pivotal role in the development of many visual search theories 2/

We explored how the TD similarity effect changes as people gain visual expertise with novel stimuli over an extended period. SS with no previous knowledge of Chinese performed a visual search task with 64 novel Chinese characters for 12 visual search sessions over 4 weeks 3/

Each target was presented x20 per session. We found a striking pattern - similarity hurt performance at the start as shown in all prior work. However! The effect disappeared by the end of session 1, and reversed for sessions 2 & onward, such that similarity helped performance! 4/

Same was observed in visual search accuracy, but only for absent trials - initially high TD similarity lead to more false alarms, but the effect quickly reversed after just 1 session of training, such that high TD similarity led to fewer false alarms. 5/

Why does this happen? We propose that greater discrimination difficulty early in the visual search task lead to the development of richer and more distinctive chunked representations of each character. This fits with some learning models like Feigenbaum and Simon (1984)'s. 6/

How can we test this further? If high TD similarity forces the development of more distinct representations, then we should expect the results to transfer to a different task performed with the same stimuli after training. This is exactly what we found in an N-back WM task 7/

These results are really surprising, but in hindsight they make sense & are an example of what Bjork calls "Desirable Difficulties". The memory system aims to make highly similar patterns more distinct from one another, so as to be better suited to support future performance. 8/

We also found that the feedback after each trial is crucial - negative feedback after false alarms becomes associated with the target representation and people slow down the next time they see it (many trials in the future). The more errors they make, the more they slow down. 9/

Here's our proposed mechanism for the surprising reversal of the target-to-distractor similarity effect. This thread glosses over many details, so for more information, I urge you to check out the paper ( https://bit.ly/2AeR3yD ) - it's quite short (main text ~4000 words). The end