Table 1.

Sensor space LSF vs HSF trial classification results. LDA classifiers were trained on scrambled image trials to differentiate between LSF and HSF image trials. CS = cluster statistic (maximum sum of cluster t-values). Classifier's generalization bias reflects SF representational dominance with negative values for CS indicating LSF dominance. Cluster numbers are in descending order of maximum cluster statistic.

SF dominance	Generalization to broadband intact	Generalization to broadband scrambled	Difference in generalization Intact minus scrambled
LSF(CS < 0)	cluster1: p <  .005, CS = -6096.12cluster2: p < .05, CS = -4652.02cluster3: p < .05, CS = -3704.29	cluster 1: p < .05, CS  =  -1834.44	cluster1: p < 0.001, CS = -7205.50cluster2: p < 0.001, CS = -1972.94cluster3: p < .005, CS = -401.12cluster4: p < .005, CS = -356.75cluster5: p < .005, CS = -241.44cluster6: p < .005, CS = -175.46cluster7: p < .005, CS = -101.59cluster8: p < .005, CS = -90.06
HSF(CS > 0)	cluster1: P < .05, CS = 2351.66	cluster1: p < .001, CS =  11,781.78 cluster2: p < .05, CS = 1801.55	n.s.