Effect of decision weights and internal noise on the growth of d′ with N

Abstract

A general finding of psychoacoustic studies is that detectability d' of a noisy signal grows less than optimally with the number N of independent observations of the signal. Competing accounts implicate internal noise common to all observations or nonoptimal decision weights given to observations. A discriminant analysis of listeners' trial-by-trial responses in a multitone level-discrimination task favored the latter account.

Object segmentation controls image reconstruction from natural scenes

“…We conclude that the sensitivity enhancement associated with rich locations on the top-down map is the outcome of sensory retuning [ 10 , 72 ] and not internal noise reduction, in line with other aspects of sensory processing [ 69 , 73 ]. Further, our data demonstrate that retuning occurs at both congruent and incongruent orientations (thick/thin blue lines in Fig 8A ).…”

“…Furthermore, this effect can be measured in the form of correlated scatter between sensitivity and top-down value, demonstrating that the performance enhancement tracks top-down information in a proportional fine-grained fashion (see S1 Text ). Sensitivity measurements by themselves, however, do not impose sufficient constraints on possible sources of improved discrimination to allow conclusions about potentially underlying mechanisms [ 69 ].…”

“…First, as discussed above, it only applies to the processing stage probed by the manipulations investigated here, which we have provisionally described as the segmented representation of the scene [ 16 , 22 , 23 ]; second, it relates to perception, not to the underlying anatomy or physiology: it is conceivable that the neural implementation may involve a default network of early visual areas [ 53 , 68 , 88 , 97 , 98 ] communicating in a fashion that may be characterized as feedback [ 99 , 100 ], although on a faster timescale than typically associated with top-down control [ 63 , 88 , 101 , 102 ]. Our results demonstrate that the visual process of reconstructing meaningful boundaries from natural scenes immediately engages such integrated extraction/segmentation perceptual module [ 29 , 53 , 68 , 88 , 92 , 101 , 103 ], the operation of which is not dependent upon attentional deployment [ 10 , 104 , 105 ] ( Fig 2G ), relies on various statistical properties of the scene ( Fig 4 ), extends over large spatial scales [ 76 , 98 ] ( Fig 6A–6G and S2D Fig ), resides within occipital cortex [ 85 ] ( Fig 7G ), and retunes its machinery to hone into the expected signal without changing its intrinsic variability [ 13 , 54 , 69 , 73 ] ( Fig 8B ). It is feasible to construct a computational support for this integrated architecture based around plausible [ 11 ] and primarily feedforward (i.e., fast) neural networks [ 106 ] ( Fig 9C–9E ).…”

Effects of relative and absolute frequency in the spectral weighting of loudness

“…The difference between the predicted and the observed performance could be attributed to either the larger amount of internal noise in listeners or lack of feedback. A cube root rule (Lutfi and Gilbertson, 2011) predicts d 0 ¼ 1.40, still higher than the observed values. Performance may be lower in this task than in other sample discrimination tasks because the components are too close to one another in frequency to be processed independently by the auditory system.…”

Dynamic Reweighting of Auditory Modulation Filters

“…by the details of the measurement task [ 43 ]). With specific relation to psychoacoustical literature using decision weights [ 44 ] (a technique related to the reverse correlation approach used here), the bandpass signatures expected of the MFB model are not directly evident through previous filter estimates [ 45 – 48 ] (we return to this issue in Discussion ). Consequently, the second goal of the present study is to measure these bandpass filters in an unconstrained manner to allow for data-driven conclusions that are largely independent of model specifics.…”