2. Perceptual Anticipation
• In sport, anticipation and perception are fundamental, related skills
- Williams et al. (2002)
Anticipation is contingent on the perception of advance visual
information from motion
- Farrow & Abernethy (2002)
• Advance Cue Utilisation
Using advance information from an opponent’s movement to predict
outcomes
- Williams (2000)
3. Expert-Novice Comparisons
• Useful for understanding nature of experts’ advantage
- Abernethy et al. (2012)
Expert players demonstrate superior anticipatory and perceptual skill
- Jackson & Mogan, (2007);
Temporal Occlusion
• Film clips edited to display variable amounts of information
Experienced players better anticipate outcomes of penalties when
clips cut before contact
- Williams & Burwitz (1993)
Standard Methodologies
4. • A means of identifying the location of beneficial advance cues
-Diaz, Fajen & Phillips (2012).
Savelsbergh et al. (2002)
• Expert goalkeepers fixate longer on the kicking leg, non-kicking leg
and ball
Novices fixate longer on the trunk, arms and hips
• Successful expert goalkeepers fixate longer on non-kicking leg
- Savelsbergh et al. (2005)
Eye Tracking
5. Present Study
• Expanded investigation to open play
Incorporated procedures with demonstrable utility
Hypotheses
H1: Experienced soccer players would demonstrate superior
anticipation
H2: Experts, intermediates and novices would differ in the areas of
the display they fixated
6. Method
• Participants were 24 males differing in soccer experience (8 in each
group)
Materials
• 48 film clips of two different football skills
32 Passes (16 high, 16 low) and 16 dribbles
Clips cut either:
o 120ms prior foot-ball contact
o 40ms prior
o At contact
o 40ms post-contact
7. Anticipation Task
At occlusion, participants stated:
1) The skill executed by the display player (Pass or Dribble)
2) Ball Direction
3) (Passes only) Ball height
11. Results
Anticipation Performance
55 65 75 85 95
Novice
Intermediate
Experienced
Direction
Height
Skill
Fig. 1
Mean percentage accuracy for each anticipatory response for
each soccer experience group.
12. Anticipation Performance cont.
N.S. difference between experience groups across the
anticipatory responses in combination (V = 0.34, F (6, 40) = 1.34, p =
0.26)
Sig. difference between experience groups in accuracy of
anticipations of skill executed (F (2, 21) = 5, MSe = 36.24, p < .05)
Experienced players were sig. > accurate than novices (p <.05)
Experience groups sig. differed in accuracy across occlusions in
combination (V = 0.66, F (8, 38) = 2.36, p <0.05)
Discriminant analysis suggested experienced & intermediate
players > accurate on clips occluded pre-contact
13. Fixation Location
• N.S. difference between experience groups in display regions fixated (Λ
= .275, F (20, 22) = 1, p = 0.5).
• Novices fixated the head (p<.05) and the shoulders (p<.05)
significantly longer than experienced players
Kicking Leg Left Arm
14. Support for H1
Experienced players more often correctly anticipated skill executed
No difference across the required anticipatory responses combined
Nor in ball direction and ball height considered separately
Contradictions of H1
Support for H2
Novices fixated the head and the shoulders longer relative to experienced
players
No difference between groups in fixation duration at the ten display regions
combined
No difference at the eight remaining display region considered separately
Contradictions of H2
Discussion
15. Discussion cont.
Limitations
• Skill of: 1) Display ‘players’ and 2) Participants
Improvements
• Alternative/additional response measures
• Incorporation of spatial occlusion trials as an adjunct procedure
Future Research
• ‘Shooting’ for goal in open play (from long and/or close range)
16. Conclusion
• Superiority of ‘experts’ extends to anticipating what opponent is
going to do
• Experts* are better able to use cues occurring earlier during
opponent’s movement
• Location of these cues also remains to be seen
Unlikely to be the head or the shoulders
Extended investigation beyond the penalty
Presenting multiple skills interchanged tested ability to anticipate
intentions
17. References
Abernethy, B., Farrow, F., Gorman, A.D., & Mann, D.L., (2012). Anticipatory
behaviour and expert performance. In Hodges N., Williams M. A. (Eds.), Skill
acquisition in sport : Research, theory and practice (2nd edition., pp. 287-306)
London ; New York; Routledge.
Farrow, D., & Abernethy, B. (2002). Can anticipatory skills be learned through implicit
video based perceptual training? Journal of Sports Sciences, 20(6), 471-485.
Jackson, R. C., & Mogan, P. (2007). Advance visual information, awareness, and
anticipation skill. Journal of Motor Behavior, 39(5), 341-351.
Savelsbergh, G. J., Williams, A. M., Kamp, J. V. D., & Ward, P. (2002). Visual search,
anticipation and expertise in soccer goalkeepers. Journal of Sports Sciences, 20(3),
279-287.
18. Savelsbergh, G. J., Van der Kamp, J., Williams, A. M., & Ward, P. (2005).
Anticipation and visual search behaviour in expert soccer
goalkeepers. Ergonomics, 48(11-14), 1686-1697.
Williams, A. M. (2000). Perceptual skill in soccer: Implications for talent
identification and development. Journal of Sports Sciences, 18(9), 737-750.
Williams, A., & Burwitz, L. (1993). Advance cue utilization in soccer. Science
and Football II, , 239-244.
Williams, A. M., Ward, P., Knowles, J. M., & Smeeton, N. J. (2002). Anticipation
skill in a real-world task: Measurement, training, and transfer in
tennis. Journal of Experimental Psychology: Applied, 8(4), 259.
Editor's Notes
Particularly in fast ball sports, such as football and tennis, due to the speed at which the ball travels, players must anticipate the direction and trajectory of the ball in advance rather than reactively.
Good anticipation requires the use advance visual information from motion patterns (Farrow & Abernethy, 2002). By advance, I mean, before ball flight becomes the source of information
So it follows that the related term…
Advance cue utilisation refers to the use of information from an opponent’s bodily orientation and posture to predict outcomes such as the direction the ball will travel and/or the height it will reach (Williams, 2000)
Standard methodologies in the field
Elite, experienced or highly skilled sports players is compared with the performance of intermediate, amateur, unskilled or novice players
Referred to as the expert-novice paradigm
Identifying underlying mechanisms – which ACU has been shown to be
Across sports experts are superior to novices
The latter of these researchers found evidence of this superiority in elite soccer players as early as age nine
These findings are intuitively appealing (obvious) and will continue to crop up as we progress;
The important thing to remember at this stage is that these comparisons (the expert-novice paradigm) are employed across studies in this area.
Another consistent feature of research in this area (another paradigm)
(Cutting clips)
That is, less or more advance and ball flight information is displayed
Not to be confused with spatial occlusion…which we’ll get into later
In soccer, Williams and Burwitz (1993) found that, relative to inexperienced players, experienced soccer players more often correctly anticipated the corner of the goal the ball would reach from a penalty kick when film clips were occluded prior foot to ball contact.
Experienced players were better at pre-contact occlusions (120ms & 40ms),
But not at foot-ball contact or 40ms post foot-ball contact
Allows for investigation of timing of extraction of pertinent visual cues
Eye movement registration technology used to determine gaze location
- Williams (2002)
Relatively objective
which promote superior anticipatory performance in experts
Savelsbergh et al. (2002) expert soccer goalkeepers (who again more accurately anticipated the direction of penalty kicks)
(classified as successful on the basis of their superior performance anticipating the direction and trajectory of penalty kicks in a film based task),
relative to their unsuccessful counterparts
Expanded investigation of advance cue utilisation to dynamic, open-play
Experienced, intermediate and novice soccer players
Temporally occluded film clips of passing and dribbling
Eye tracker recorded participants’ fixations
Relative to intermediates and novices, more often correctly anticipating: 1) the skill executed by the display player, 2) the direction the ball would travel, and 3) the height the ball would reach.
Experts, intermediates and novices would differ in the time they spent fixating predetermined areas of the display, during the anticipation task
Participants were assigned to the experienced soccer player group if they stated that they currently played organised (at least Saturday or Sunday league level) soccer; to the intermediate group if they indicated that they had played organised soccer but no longer played and had not for a number of years (however, most if not all in this group remained interested spectators); and to the novice group if they stated never having played organised soccer and having no interest in watching soccer.
Differing
(one left-footed and one right-footed) in soccer experience: Experienced (n=8, Mean age = 22.63, SD = 2.13, Range = 6), Intermediate (n=8, Mean age = 22.25, SD = 2.38, Range = 7), and Novice (n=8, Mean age = 24.5, SD = 9.9, Range = 28)
performed individually by two amateur players were created by the researcher.
Four (an equal number) of each skill were directed toward the left, central left, central right or right
to be differentially occluded (the films ended and were replaced in the display by an entirely black screen)
In combined form, these temporal occlusion instances are identical to those of Williams and Burwitz (1993) and a number of researchers have employed at least one of these instances (for example, Ward and Williams [2003] occluded clips at 120ms prior to foot-ball contact). Each clip lasted between two and three seconds total.
At the instance of occlusion (at which time, the display reverted to an entirely blank screen)
A still image of complete blackness was inserted into the playback order before the first clip, in between each successive clip and after the final clip.
4 practice clips
44 ‘proper’
The procedure for the experiment proper was the same as for the practice detailed above, except that instead of four clips, the remaining 44 (30 passes; 15 aerial, 15 flat and 14 dribbles; 11 at each occlusion
Aerial (High) pass, Centre-Left ….40ms post foot-ball contact
Low pass, Left 120ms prior
The system works by illuminating the eye with infrared (IR) LED at an angle from an IR sensitive camera. The eye and face reflect this illumination but the pupil will absorb the IR light and appear as a high contrast dark ellipse. Image-analysis software determines the location of the centre of the pupil and this is mapped to gaze position via an algorithm. Images of the eye are analysed in real-time, with small head movements compensated for by tracking of the corneal reflex. This system allows for contact-free measurement of eye movements and is accurate to within 1° of visual angle.
Calibration
During calibration, the system displays an all-white reference circle with a minute red dot located centrally within. Upon the participant pressing the space bar, this circle moves in a random order to each corner of the display, twice. The participant was instructed to follow the reference circle with their eyes, whilst keeping their head still. Once the circle had moved to each corner twice, the calibration attempt is complete and a display of the participant’s accuracy (a cross placed where the eye fixated in relation to the circle at each corner), along with precise details of average degrees of visual angle deviated from the target, appeared on the experimenter’s control system. At this point, the experimenter used this information to decide whether to recalibrate, or to accept the calibration and continue with the procedure. Calibrations were accepted if average deviation was within 1 degree of visual angle (horizontally and vertically) of the target. Recalibration was undertaken for average deviations of more than 1 degree of visual angle (either horizontal or vertical).
Using Pilla’s Trace statistic, there was no significant difference between experience groups across the anticipation dependent variables in combination (V = 0.34, F (6, 40) = 1.34, p = 0.26). Follow up univariate analyses of variance (ANOVAs) indicated that there was a significant difference between experience groups in percentage of correct predictions of skill executed (F (2, 21) = 5, MSe = 36.24, p < .05), but not in percentage of correct predictions of ball height (F (2, 21) = 2.45, MSe = 165.34, p = 0.11), or in percentage of correct predictions of ball direction (F (2, 21) = 1.48, MSe = 170.95, p = 0.25). Post-hoc multiple comparisons, with Bonferroni’s alpha adjustment applied, revealed that experienced players were significantly more accurate than novices (but not intermediates [p = 0.88]), in anticipating the skill executed by the display player (p <.05).
To determine whether experience groups significantly differed in terms of their anticipation accuracy across the different occlusion instances a one-way MANOVA, with soccer experience as the fixed factor and the percentage accuracy scores at each of the four occlusion instances as the dependents was run. Box’s Test was not significant F (20, 1583) = 1.22, p = 0.23. Furthermore, Levene’s tests were not significant across the occlusion instances: F (2, 21) = .76, p = 0.48) for 120ms prior to foot-ball contact; (F (2, 21) = 0.58, p = 0.57) for 40ms prior; (F (2, 21) = 1.44, p = 0.26 for at foot-ball contact; and (F (2, 21) = 0.23, p = 0 .8 for 40ms post-contact. Using Pilla’s Trace statistic, soccer experience groups significantly differed in the accuracy of their anticipations across the different temporal occlusion instances (V = 0.66, F (8, 38) = 2.36, p <0.05.) None of the univariate F tests attained significance: F (2, 21) = 1.8, MSe = 46.19, p = 0.19 for accuracy on clips occluded 120 ms prior; F (2, 21) = 0.7, MSe = 25.11, p = 0.51 for accuracy on clips occluded 40ms prior; F (2, 21) = .68, MSe = 60.52, p = 0.52 for accuracy on clips occluded at foot-ball contact; and F (2, 21) = 1.28, MSe = 43.12, p = 0.3 for accuracy on clips occluded 40ms post foot-ball contact. This absence of a difference between groups in univariate terms, in conjunction with the significant MANOVA result, suggested a discriminant function analysis should be run to establish whether a linear combination of the dependents would differentiate soccer experience groups.
A discriminant function analysis, with soccer experience as the grouping variable and percentage accuracy on clips occlude at the four instances as the independents, was run. Two discriminant functions were obtained. The first function explained 93.6% of the variance, canonical R2 = 0.58; and the second explained 6.4%, canonical R2 = 0.09. In combination, these functions significantly differentiated the soccer experience groups (Λ = 0.39, χ² (8) = 18.57, p < 0.05). Removal of the first function revealed that the second alone did not significantly differentiate the groups (Λ = 0.91, χ² (3) = 1.75, p = 0.63). Correlations between the occlusion instances and the functions showed that accuracy on clips occluded 40ms post foot-ball contact had a weak, negative loading on the first function (r = -0.29) and was similarly correlated with the second (r = -0.26). Accuracy on clips occluded 40ms prior had a weak, positive loading on the first function (r = 0.16) and a moderate-to-strong, positive loading on the second (r = 0.57). Accuracy on clips occluded 120ms prior had a moderate, positive correlation with the first function (r = 0.34) and a similarly moderate, but negative correlation with the second (r = -0.35). Accuracy on clips occluded at contact had a weak, negative correlation with the first function (r = -0.21) and a similarly weak, but positive correlation with the first (r = 0.23). Inspection of the discriminant function plot revealed that the first function differentiated intermediate and experienced players from novices, and the second, to a lesser extent, differentiated experienced players and novices from intermediates.
There was no significant difference between experience groups in duration of visual fixations across the different regions of the display in combination (Λ = .275, F (20, 22) = 1, p = 0.5). However, follow up ANOVAs indicated there was a significant difference between groups in the total duration of visual fixations (across trials) on the head of the display player (F (2, 20) = 5.55, MSe = 4478931.94, p < 0.05) and the shoulders of the display player (F (2, 20) = 4.21, MSe = 6043349.22, p < 0.05). Post hoc comparisons (with Bonferoni’s correction applied) revealed that novices fixated the head significantly longer than experienced players (p < 0.05), but not intermediates (p = 0.09).Novices also novices fixated the shoulders longer across clips than experts (p < 0.05). Otherwise, there were no significant differences between groups in terms of time spent fixating the different display regions.
The visual fixation dependent variable was the total duration (ms) of visual fixations on nine different predetermined regions of the body of the display player across all clips displayed. These nine regions were: the head, shoulders, torso, arms, hips, kicking leg, kicking foot, non-kicking leg, and the non-kicking foot. The ball was also selected as an area of the display in which fixations were to be measured. Fixations not in these locations were recorded as ‘unclassified.’ SMI iView X™ gaze analysis software automatically registers each fixation and its duration. The researcher reviewed focus map output (where the fixated area is highlighted and the rest of the display appears darkened) for each clip in order to determine fixation location. Thus, these gaze analysis features in combination allowed for summation of duration of fixations in the predetermined postural regions.
Relates to realism of stimuli
Relates to magnitude of observed skill based differences
In addition to ensuring a high degree of realism in the presented stimuli, the use of alternative response measures represents another means of improving this research should a partial replication be undertaken. More specifically, instead of verbalising anticipatory responses, participants could be required to respond using a joystick as in Savlesbergh et al.’s experiments (2002; 2005), or a movement based response system could be incorporated as in Williams and Davids (1998; wherein participants’ moved right, left, forward, or backward to simulate intercepting a pass, or right or left to intercept a dribble, and pressure pads were used to obtain timed information regarding, initiation and duration of movement and reaction timings). The reasons alternative response measures are suggested are twofold: 1) incorporation of these would allow for differences in reaction time as well as accuracy to be ascertained, and 2) with movement based responses particularly, the coupling of perception and the execution of movement(s) (perception-action coupling) is more representative of actual sport performance (and perhaps consequently, is thought to better discriminate between skill levels [Dicks, Button & Davids, 2010).
Another feature that would enhance the validity of the current experiment as a test of advance cue utilisation in open play soccer is the incorporation of spatial occlusion trials as an adjunct procedure. Indeed, the systematic occluding of display regions (normally body segments, racquets or bats), and the subsequent examination of decreases in performance allows for inferences to be made regarding the importance of regions as advanced cues; this procedure has been employed in combination with the temporal occlusion approach to determine the specific location of beneficial advance information utilised by expert performers (Abernethy et al., 2012). Moreover, the sheer diversity that characterises open play soccer is an almost limitless source of potential directions for further research. For example, ‘shooting’ for goal in open play (from long and/or close range) could be investigated through employment of the filmed temporal occlusion paradigm, as could volleys and crosses in to the penalty area. These different skills could be presented in different areas of the pitch, at different angles in relation to the goal, and so on. Furthermore, while these different skills could be analysed separately, presenting them interspersed within the same task, as passes and dribbles were in the experiment reported here, is not only more representative of actual soccer play but also allows for analysis of whether and how the execution of different skills can be anticipated.
when there is uncertainty in this respect.
*and, to a lesser degree, intermediates.
