These data code category "X" as numerical value 1.0 and category "O" as numerical value 2.0. For the final test items, "equally X or O" is coded as numerical value 1.5. These data come from a version of the experiment in which the test trials had five response options (instead of three): Very Probably X (=1.0), Probably X (=1.25), Equally X or O (=1.5), Probably O (=1.75), and Very Probably O (=2.0). Data for training cases are from last block of training (trials 13-16). These data are from 76 participants.

Rectangle Height	Line Segment Position
	1	2	3	4	5	6
6	Test case: Correct = ? Mean = 1.21 SE = 0.033	.	.	.	.	.
5	Test case: Correct = ? Mean = 1.18 SE = 0.029	.	.	.	Training case: Correct = 2 Mean = 1.92 SE = 0.031	.
4	.	.	.	(prototype)	.	.
3	Training case: Correct = 1 Mean = 1.09 SE = 0.033	.	Training case: Correct = 2 Mean = 1.82 SE = 0.045	.	.	.
2	.	(prototype)	.	.	.	.
1	.	.	Training case: Correct = 1 Mean = 1.25 SE = 0.050	.	Test case: Correct = ? Mean = 1.61 SE = 0.038	Test case: Correct = ? Mean = 1.58 SE = 0.038

t-tests confirm that all the means are significantly different from 1.50. That is, there was a significant bias toward one category or the other for every case.

Which models can fit these data? These group average data are not fit by the prototype model described in the homework, nor by the exemplar model described in the homework. The rule-and-exception model does fit these data. There are more elaborate versions of the prototype and exemplar models, however, that could fit these data. Your individual data might not match these average data, so different models, or none of them, might fit your data.