Feature Search On Quantitative Dimension

Treisman and Gelade (1980) have put forward a Feature-Integration Theory of Attention (FIT). Visual search is an important part of FIT. It includes feature search and conjunction search. Feature search is at the preattentive processing stage and belongs to parallel processing. Conjunction search is at the attentive processing stage on the basis of feature search and attributes to serial processing. Feature search mainly contains searches on quantitative dimension and color dimension. Feature search on quantitative dimension plays an important role.Two questions on quantitative dimension that are not resolved at present are put forward according to some theories and researchers' results. The questions are as follows: (1) Main factor that decides search slopes is determined. (2) Critical point that differentiates parallel search and serial search is fixed from physical property of stimuli.After a ratio(C) is drawn into, three hypotheses are given in order to solve the two questions. The ratio C is difference between target value(T) and distractors values(D) divided by distractors values. Three hypotheses are as follows: (1) A function of slopes to C absolute values is monotone decreasing function. (2) Slopes are equal when C values are same. (3) If C absolute value is bigger than C_p absolute value, searching for a target is parallel search, or is serial search. Moreover, forms of stimuli do not affect truth of these hypotheses.Two experiments in the first research examine the first and second hypothesis by circle and triangle stimuli. When target values are bigger than distractors values, the two experiments have found that a function of slopes to C values is monotone decreasing function, that slopes are identical under same C values, and that there is no significant difference between circles stimulus and triangles stimulus under same C values.The second research has also got that a function of slopes to C values is monotone decreasing function when target circles are bigger than distractors circles. The critical point between parallel and serial search is C_p=0.497. Searching for a target is parallel when C≥0.497. That is, slope isequal to or smaller than lOmsec/item. Searching for a target is serial when C<0.497. That is, slope is bigger than lOmsec/item.Target circles are smaller than distractors circles in the third research. The research has attained that a function of slopes to C values is monotone increasing function. The critical point between parallel and serial search is Cp=-0.57. Searches targets are serial when O-0.57. That is, slopes are bigger than lOmsec/item. Searches targets are parallel when C<-0.57. That is, slopes are equal to or smaller than lOmsec/itemFinally, a C value is randomly chosen within parallel search ranges of the second and third researches in order to test whether searches within the range are parallel search. Two slopes that experiment 4 has got before and after target and distractors are interchanged have no significant difference. Moreover, searches are symmetry when C values belong to parallel search ranges.Four researches and discussion are summarized. Main results are as follows: (1) Slopes are explained by "relative difference" concept more reasonable than "absolute difference" concept. (2) Hit rates and false alarm rates decrease when C values are small and set sizes are large (12 items). Signal Detection Theory and capacity of short-term memory (7±2) explaine the phenomenon. (3) The ratios of absent-target slopes to present-target slopes are nearly 2.0.