Psychophysics

W hen som ething m oves, two crucial param eters o f interest are its direction o f m ovem ent, and its speed. The ability to register both o f these attributes has been studied m any tim es. The presence o f cells in prim ate extra-striate cortex that are responsive to wide-field, or global patterns o f m otion has led to renew ed interest in these topics. The m edial tem poral (M T) and m edial superior tem poral (M ST)

areas contain m otion-sensitive cells with m uch larger receptive fields than those o f V I , w hich p ro jects to them (M aunsell & N ew som e, 1987). I f these m echanism s rep resent a fu rth er stage o f m otion analysis, their speed and directional tuning com petence may be different to that m easured by previous investigators. The follow ing section deals w ith w hat is currently know n about speed and direction capabilities, as m easured by behavioural techniques.

1.2.5.1 Speed and Direction in Local Mechanisms

In the first instance the visual scene is analysed for the presence o f m otion in discrete regions. This ‘lo ca l’ analysis introduces a tension betw een com peting requirem ents if estim ates o f both speed and direction param eters are required from the sam e m echanism . To sense m otion it is necessary for a system to som ehow register the tim e taken to traverse a distance. The ability to obtain an accurate local estim ate o f speed depends on keeping the distance and tim e values in the equation low, avoiding averaging. Y et the accuracy o f a direction estim ate is im proved if these values are increased (G rzyw acz et al., 1994). C alculations of the error involved in local speed discrim ination by hum an observers range from 30% to 100%, indicating that speed is not initially com puted w ith great accuracy, (V aina et al., 1990; Bravo & W atam aniuk, 1991). W hen local signals are com bined, however, precision increases greatly, with discrim ination errors as low as 5% being reported (M cK ee et al., 1986; V aina et al., 1990; B ravo & W atam aniuk, 1991). These estim ates w ere obtained using random dot displays w ith isotropic (circular) elem ents, so it is certain that subjects w ere relying solely on their perception o f m otion direction, rather than using an artifactual orientation cue. The high errors in local m otion speed estim ation m ight lead us to conclude that accurate local speed estim ates have been sacrificed in favour o f directional tuning.

Y et estim ates o f the directional bandw idths o f m otion m echanism s indicates that these too are broad, in the region o f 130 degrees, (Ball & Sekuler, 1980; Ball et al., 1983). H ow ever, there are tw o pro b lem s w ith prev io u s estim ates o f directional tuning. R eported values for bandw idth have been estim ated from

experim ents which have been designed to gauge how great the angular distance betw een signals is required to be in order that the m echanism s responding to them do not interact with each other, (e.g. Levinson & Sekuler, 1979; M ather & M oulden, 1980; Ball & Sekuler, 1979; Ball & Sekuler, 1980; Ball & Sekuler, 1981). These experim ents found that directions o f m otion had to be betw een 120 degrees and 150 degrees apart for independence betw een m echanism s to be evident. T hese figures have then been interpreted directly as bandw idths o f directional tuning. How ever, a principled estim ate o f bandw idth is obtained at the h a lf height o f the m echanism s response curve. C onsider figure 1.2 which shows a Gaussian profile representing the response profile o f a motion detector sensitive to a finite range o f directions. Conventionally the range o f directions to w hich it responds is quoted at h alf its m axim um response, hence ‘half-height bandw idth’. This definition is useful as an attem pt to m easure its w idth at the base would be problem atic due to the theoretically infinite nature o f the decay o f the G aussian function. In addition, low -level noise w ould be less likely to contam inate width estim ates m ade at half height. Estim ates m ade for half-height bandw idth have given rise to m uch sm aller figures o f 70-90 degrees (Raym ond,

1993). The point o f independence betw een m echanism s is a region o f the response curves that may slope very gently, and give rise to m isleadingly large and variable estim ates. Secondly, the design o f the random dot stim uli used in som e o f these experim ents did not preclude subjects from integrating across m any elem ents, hence p ossibly p ro d ucing b a n d w id th estim ates o f global directio n detectio n m echanism s. D irectional tu n in g , as m easured by unit recording techniques in m onkey, results in bandw idths o f circa 40 degrees for V I cells (De V alois, Yund & H epler, 1982), and circa 80 degrees for M T cells (Fellem an & Kaas, 1983).

R esponse profile of motion detector

m otio n d ire c tio n (d e g re e s )

4 0 5 0 6 0

Figure 1.2. T h e half-height b an d w id th o f a m otion d e te c to r with a Gaussian tunin g profile. The detec tor responds m a xim a lly when stim ulated by a motion at 0 degrees. T h e horizontal line is po sitio n ed at h a l f this peak. T h e w idth o f the m e c h a n is m is conventionally taken where the horizontal line intersects the Gaussian profile and can be estim ated from the x axis by projecting vertically from the intersection points. In this ex am p le the half-height bandw idth is from - 1 8 degre es to 18 degrees, i.e. 36 degrees. Alternatively the b andw idth can be c o m p u ted m a th em atically by taking the inverse o f the Gaussian form ula at the intersection points.

1.2.5.2 Speed an d D irection in Global M echanism s

M e a s u r i n g the d ir e c t i o n a l tu n i n g b a n d w i d t h s o f c o m p l e x m o t i o n d e t e c t o r s re q u ire s that stim uli o b lig e j u d g e m e n t s to be m a d e u sin g global d e te c to rs rath er than local ones. E v id e n c e has b een a c c u m u la tin g fr o m p s y c h o p h y s ic s (R e g a n & B e v e rle y , 1978; M o r r o n e , B u rr & V a in a , 1995; B u rr, M o r r o n e & V a in a , 1998; S n o w d e n & M i l n e , 1 9 9 7 ; M e e s e & H a r r is , 2 0 0 0 ) a n d e l e c t r o - p h y s i o l o g y ( T a n a k a & Saito, 1989; D u ffy & W u rtz , 1991 ; G r a z ia n o , A n d e rs e n & S n o w d e n ,