This study investigates the capacity and resolution of visual working memory. The authors propose that working memory stores a limited number of discrete, fixed-resolution representations, rather than a flexible resource pool. They tested this hypothesis using a colour recall task, where participants reported the colour of a probed item by clicking on a colour wheel. The results showed that when presented with more than a few simple objects, participants stored a high-resolution representation of a subset of the objects and retained no information about the others. The precision of the representation varied within a narrow range, consistent with a small set of discrete, fixed-resolution representations. The study used a mixture model to separate the probability of the probed item being stored in memory (Pm) and the precision of the representation (s.d.). Experiment 1 showed that Pm was approximately twice as great at set size 3 as at set size 6, while s.d. did not vary significantly. A simple fixed-resolution model provided an excellent fit to the data, ruling out models where all items are stored with resolution dependent on the number of items. Further experiments supported the fixed-resolution model, showing that precision improved when set size was reduced below the number of available slots. The results also supported the 'slots+averaging' model, where precision increases with the number of slots. However, the 'slots+resources' model predicted larger differences in precision between valid, neutral, and invalid trials, which was not observed. The study also tested the 'slots+resources' and 'slots+averaging' models using a line in the sample array to cue one of four coloured squares. The results supported the 'slots+averaging' model, showing that precision was higher for valid trials than neutral trials, but not for invalid trials. The study concludes that visual working memory stores a small number of discrete, fixed-resolution representations, with precision limited by the number of slots. This model provides a quantitative account of memory performance across a broad range of experimental manipulations. The results suggest that the capacity of visual working memory is set by both information load and the number of objects, and that the process of creating durable memory representations involves an all-or-none step. The findings have implications for understanding the mechanisms of visual working memory and its role in cognitive functions.This study investigates the capacity and resolution of visual working memory. The authors propose that working memory stores a limited number of discrete, fixed-resolution representations, rather than a flexible resource pool. They tested this hypothesis using a colour recall task, where participants reported the colour of a probed item by clicking on a colour wheel. The results showed that when presented with more than a few simple objects, participants stored a high-resolution representation of a subset of the objects and retained no information about the others. The precision of the representation varied within a narrow range, consistent with a small set of discrete, fixed-resolution representations. The study used a mixture model to separate the probability of the probed item being stored in memory (Pm) and the precision of the representation (s.d.). Experiment 1 showed that Pm was approximately twice as great at set size 3 as at set size 6, while s.d. did not vary significantly. A simple fixed-resolution model provided an excellent fit to the data, ruling out models where all items are stored with resolution dependent on the number of items. Further experiments supported the fixed-resolution model, showing that precision improved when set size was reduced below the number of available slots. The results also supported the 'slots+averaging' model, where precision increases with the number of slots. However, the 'slots+resources' model predicted larger differences in precision between valid, neutral, and invalid trials, which was not observed. The study also tested the 'slots+resources' and 'slots+averaging' models using a line in the sample array to cue one of four coloured squares. The results supported the 'slots+averaging' model, showing that precision was higher for valid trials than neutral trials, but not for invalid trials. The study concludes that visual working memory stores a small number of discrete, fixed-resolution representations, with precision limited by the number of slots. This model provides a quantitative account of memory performance across a broad range of experimental manipulations. The results suggest that the capacity of visual working memory is set by both information load and the number of objects, and that the process of creating durable memory representations involves an all-or-none step. The findings have implications for understanding the mechanisms of visual working memory and its role in cognitive functions.