PS3060: Perception and Action
Term II,  MONDAY 10 – 12 am  (Room 128 Wolfson)

Lecture 4: Cricket: how to hit a ball that can’t be seen

Course co-ordinator: Johannes M. Zanker, j.zanker@rhul.ac.uk, (Room 218)

Topics Lecture 4:

The visuomotor problem of intercepting moving objects

biological relevance: predators have to intersect prey (sea eagle); in many instances animals have to catch mating partners (flies)
in humans intercepting moving objects is required by a wide range of very common activities like reaching/grasping, catching, hitting with tool

experimental analysis starts with simplified situations 

basic geometry of interception (Lishman 1981)

similarity to collision avoidance : collision achievement !

however : extensive debate whether tau theory is a sufficient or necessary part of understanding human behaviour


corresponding constructions for other trajectories of object relative to observer, and for collision of 2 objects (Bootsma and Oudejans 1993)


however: extensive debate whether tau theory is a sufficient or necessary to understand human behaviour
  •   generalisations would generate typical errors, if no additional rules are put into place
  •   focus on tau would be ignoring other useful cues (Tresilian 1999) (Wann 1996)

>> object interception good example to study visuomotor mechanisms

Motion in depth: monocular and binocular cues

in a more general case, variable direction of approaching object, the geometry is less simple:
time and distance of intersecting image plane are crucial variables
monocular information is ambiguous: optical estimates lateral distance of interception from observation point, for instance, depends on objects size (Regan and Kashal 1994)
 
the binocular geometry of interception trajectories provides additional information: estimates of object size, cues about direction of motion from disparity changes in the two eyes (Regan et al. 1995)

monocular and binocular information can be used in a similar way to estimate the timing of an interception event  (Regan and Gray 2000)

simulated approaches suggest that various cues necessary to construct such estimates can be extracted by human observers independent of each other, if required (Regan and Kaushal 1994) (Regan and Gray 2000)

psychophysical experiments further demonstrate:

Hitting the spider paradigm

is the position of a target specified in the same way in the perception and action systems ?

a typical fast goal directed action is used to study potential differences between perception & action
peculiar properties, such as misperceived object motion in presence of background flow (Brenner 1991), can be exploited to compare characteristic patterns
>> experimental question: do we find changes of action patterns similar to those found in perception?
 

basic paradigm (Smeets and Brenner 1995)
  • spider walking across screen, structured background static/moving independently
  • hit spider on protective screen with perspex rod 
  • compare the speed/final position of two successively presented spiders

Perceptual matches

  • speed comparison for static background is very accurate
  • perceived speed is enhanced/reduced by background motion in opposite/same direction
  • perceived position is very accurate for all speeds
  • perceived position is unaffected by background motion

Hand movements

  • RT (reaction time) reduced for faster spiders
  • for slow spiders: RT reduced for opposite background motion, increased for background motion in same direction as spider (i.e. treated like faster/slower spiders)
  • MT (movement time) decreases for faster spiders (i.e. average hitting speed grows with spiders velocity)
  • MT is reduced for opposite background motion, increased for background motion in same direction as spider (i.e. treated like faster/slower spiders)
  • overall direction of hand trajectory varies systematically with spider speed (home in on spider position at screen impact) > lateral position converted in ‘apparent speed’
  • direction estimates are unaffected by background motion

Hitting and seeing moving targets

general conclusions:

>> in this paradigm, perception and action seems to be based on same information processing mechanisms !
further support by observations on changes of hitting trajectories when the spiders movement is irregular (Brenner and Smeets 1997) (Brenner et al. 1998)

Response to stimulus perturbations during execution of hand movements

perturbations of spider (background) trajectory after onset of hand movement:

  • brief jump of spider position >> lateral correction to meet target after about 110 msec
  • background movement with static spider (and controls of moving spiders with static background) >> no effect of background movement on final lateral position

perceptual measurements lead to corresponding results

these results have been interpreted as evidence that after onset of hand movement subjects respond with lateral trajectory adjustments to changes in position rather than to changes in velocity (Brenner and Smeets 1997)

further evidence that the acceleration of the hand is continuously adjusted on basis of target speed with a delay of 200 msec (Brenner et al. 1998)

Catching of tennis balls

Is it necessary to see the complete path of a moving target to predict interception point & catch? (Whiting & Sharp 1974)

best performance for 285 msec darkness, worse for shorter and for longer durations of blackout before catch
>> optimum time window to predict flight path 250 –300 msec before impact
>> minimise duration of storage in memory and meet minimum processing time to extract flight path information

this observation would be accounted for by the general suggestion that 300 msec TTC are critical to initiate and execute a standardised motor pattern (Lee 1980)

Punching a dropping ball – the quick and dirty estimate of TTC

a ball dropping towards the ground accelerates at constant rate (G) and therefore violates the assumption of constant speed that is necessary for using tau (tau = theta/(d.theta/d.t)) as TTC estimate

however, close to impact tau converges towards the veridical TTC value, and would provide a reasonably good estimate far any time shorter than 250 msec before contact

knee and elbow kinematics were measured in subjects initiating a jump to punch a ball dropping from various heights (Lee et al. 1983)

this result can be interpreted as strategy to use a critical tau value to initiate the jump and later fine-tune the punch movement using more accurate information

Critical test: the deflated ball

so far just observation of correlated events >> what would be critical prediction to prove that tau is used?
the experimental manipulation of the object's expansion rate should change response timing!
grasping a ball which is changing its size
(Savelsbergh et al. 1991)

inflating ball leads to reduction of relative expansion, as compared to constant size, with identical speed

how does this affect kinematics of grasping?

 
hand aperture (thumb-index finger) initially like that for large ball, later close to that for small ball

maximum closing velocity:

  • large : -46 msec
  • small : -40 msec
  • deflating: -24 msec
    		•

    the process that initiates closing the aperture follows the prediction of estimating TTC by tau:
    significantly shorter for deflating ball (reduced relative expansion rate)

    additional support for this conclusion from experiments with limited viewing time :
    different critical period for deflating ball (Savelsbergh et al. 1993)

    Run & catch: the outfielder’s problem

    problem of catching a ball is aggravated when the ball is not on intersection trajectory within catchers reach (see above for geometry of general solution)
    the outfielder in baseball or cricket has to run for a ball flying on a ballistic trajectory
    a simple geometric strategy seems to help: the outfielder is moving on a path that keeps the perceived trajectory of the ball straight
    (McBeath and Shaffer 1995)


    this will bring the fielder into the path of the ball when the fielder reaches the point of landing

    video analysis of professional fielders show that they follow a path at variable speed, keep watching the ball and catch on the run
    – just as predicted by this hypothesis !

    other, simplified strategies may apply for cases when the ball flies in the fielder’s direction:
    moving forwards and backwards to catch the ball, following a simple geometrical rule
    - keep the second temporal derivative of elevation constant (McLeod and Dienes 1993)

    Finally, what is the batsman’ problem?

    fast bowling is dangerous !
    HRH Frederick Louis, Prince of Wales, was killed by cricket ball in 1751
    •  where does the ball move after bouncing ??
    •  what is the remaining time (depending on speed   of the ball and the point where it hits the ground) ??

      •  

       

      short-pitched, fast ball: (Regan. 1992)

      • 90 mph (40 m/sec)
      • 30 ft (10 m) from batsman
      • 250 msec to coordinate stroke !

    to predict these variables requires outstanding precision !!!!!!!!!!!

    so what are the sensory cues actually used ?

    A few myths, a few facts

    Victor Trumper,
    London Oval, 1902

    impact between bat and ball in spatial and   temporal ‘window’:
    10 cm, 5 msec (Regan. 1992)

    Saw it ? How fast ?

    Reaction times ? Colours and light levels ? Body movement mimicry ?

    Some suggestions

    what are the visual cues (simple optical parameters in Gibsonean sense) that batsmen can use ? (Regan 1992)

    >> the visual system is pushed to its limits !!!!!!!

    So what do batsmen do ?


    a geometrical analysis of fast bowling in the vertical plane may suggest simplified strategies to estimate time and level of arrival (t1, y) for predictive judgements (Land and McLeod 2000)

    foveal region provides best visual information
    > eye movements can reveal visuomotor strategies

    the experiment:

      eye and head angles in relative coordinates
      >> gaze direction and ball direction (in space coordinates, from head level)

       

       

      and this is what bowlers do:

    • rest on point of delivery
    • early saccade (loosing ball) : close to subsequent bouncing point (counter-rotation of head and eyes)
    • capture of ball in centre of gaze, tracking in lower regions

     
    effects of delivery length and batsman skills on saccacde pattern : earlier, fewer saccades by professional

    ability to combine pursuit and saccades tracking mainly after bounce, for not more than 200 msec, before that not keeping eye on the ball !

    unambiguous information only after bounce: the pre-bounce height and distance can be mapped on the post-bounce height and distance, knowing the surface properties of the ground (acquired through a number of initial defensive shots, adjusted through tracking) >> mapping a such requires experience

    key reading:

    comprehensive reference and reading list:

    some study questions

    download lecture handout

    back to index page
    last update 23/02/2004
    Johannes M. Zanker