Effective instruction may be crucial to the pursuit of optimal sporting performance. The most significant role of the physical education teacher or the coach is to give information about the skills' execution in the form of feedback (Hodges and Franks, 2002) and has been found to be a key tool in improving and learning motor skills (Schmidt and Wrisberg, 2004). According to the cognitive approach the role of instructions and criticism on performance is a crucial factor for learning (Wulf and Shea, 2004) however, the ecological (Gibson, 1979) and dynamical systems approach (Kelso, 1981; Stergiou, Harbourne and Cavanaugh, 2006) of performance and learning support that information about the movement from an external source feedback is a second order constraint.

Children begin to form impressions about their own self-worth based on the types of experiences they have and the nature of the feedback they get about their performance. If children are to feel competent, teachers must give them appropriate information about their performance. It is not enough merely to praise them for trying. Teachers must be selective in providing reinforcement and be certain that the behaviour of a child is appropriate for a particular reinforcement. Many researchers attempted to find the most appropriate methods of providing information through feedback to refine and develop motor skills (Salmoni et al., 1984). It is important to realize that this information can be acquired through many different methods, not all of which are as effective as each other (Amorose and Weiss, 1998; Williams and Hodges, 2005). Providing feedback in the form of verbal cueing facilitates the performance of the task by verbally indicating vital form characteristics (Landin, 1996), enhances attention and provides additional information that may not be available through visual observation. (Janelle et al., 2003). Lee et al., 1993 suggested that instructors' feedback is typically verbal and in the form of positive, non-specific evaluative statements. Providing verbal cues about errors and corrections is useful for learners especially for the beginners (Kernodle and Carlton, 1992). The provision of encouragement with feedback that will help improve a skill (corrective feedback or criticism) may help the child improve and also believe the idea that the child can do better and improve the self confidence.

However, less is known about the effectiveness of feedback in skills of different difficulty level (More and Franks, 1996; Hughes and Franks, 1997). Kernodle and Carlton, 1992 supported that when feedback provided contains error and correction cues is more useful for the difficult tasks. Schmidt and Lee, 1998 proposed that more research is needed for examining the relation of skills' difficulty with feedback effectiveness.

A mediating factor between the presentation of the instructions by the coach and the performance of the skill by the player might be the cognitive process of self confidence (Escarti and Guzman, 1999). The most powerful source of self-confidence is mastery of a skill (Bandura, 1977; 1997; Harter, 1978; Vealey et al., 1998). Allen and Howe, 1998 showed that self-confidence of the athletes is determined by coach feedback. Feedback can be accepted as a reward when there is a correct execution and this may increase self confidence levels or as a criticism when errors of the execution are corrected and the level of self confidence is decreased. Smith et al., 1995 asserted that feedback either in the form of error correction or in the form of praise and criticism can have a significant effect on young athletes' psychology and self confidence. Feltz, 1988 argued that it is still not clear whether the guidance and feedback of a coach can have an effect on an athlete's self-confidence. Lyster and Ranta, 1997 attempted to find what types of error treatments encourage learners and the results have been found to be quite complicated. Research on instructor feedback and student uptake does not yield conclusive claims and more research is needed.

Many researchers investigated different methods of instruction that improve learning in laboratory settings (Vickers et al., 1999) in classroom settings (Scheeler and Lee, 2002) or in applied settings (Goode and Magill, 1986) However, there is not much empirical evidence for the effect of instructors' corrective feedback on skills' learning and self confidence for different type of skills (Franks, 1997; Hughes and Franks, 1997; More and Franks, 1996) including sport related cognitive complex solving tasks (McCullagh and Little, 1990; Sanchez and Bambouras, 2006; Silverman, 1994). Rink et al., 1996 proposed that the lack of empirical evidence to support any one approach to the teaching-learning process over another precludes the efficacy of suggesting a 'model' profile for coaches' pedagogical content interventions. It is important for coaches to know how these different sources of feedback work both alone and in conjunction with other instructional techniques to improve learning of different complexity skills. Scheeler et al., 2004 mentioned that determining and interpreting the impact that different types of feedback have on performance has been difficult because there are number of complex theoretical processes of mechanisms involved. Williams and Hodges (2005) added that it might need further study before a complete understanding of its nature and significance is possible. Bunker, 1991 suggested that children acquire self-confidence and self-esteem as a result of successful experience.

This experiment was designed to investigate how different types of corrective feedback (positive and correction cues, error cues or a combination) improve learning of a skill and alter self confidence of the participants. An additional purpose was to investigate whether these corrective feedback methods have the same effects in easy or difficult skills. In previous experiments (Tzetzis and Votsis, 2006), the effect of feedback may have been confounded with the level of difficulty of the skills, thus this experiment manipulates the level of difficulty to see whether feedback differentially affects outcome and self confidence when the skill difficulty varies. The purpose was to investigate the effect of three different treatment conditions (corrective feedback), the effect of time and the effect of difficulty level as well as their interaction on retention of the outcome and self-confidence, of two badminton skills with different difficulty level, for youth participants. The relationship of the participants' self confidence by their outcome scores was also investigated. Since participants are novices and corrective feedback as well as error identification feedback are both necessary in directing, correcting and motivating them in practice, it was hypothesized that the combined method of positive feedback and instructional cues on execution as well as error cues would have the best results, improving the outcome and self confidence scores of both skills, across time.

Description of the skills
The two skills were: a) forehand-clear: a high return stroke on the dominant side of the body that carries the shuttlecock deep in to the backcourt (low difficulty) and b) backhand-clear: a high return stroke on the no dominant side of the body that carries the shuttlecock deep in to the backcourt (high difficulty). The two badminton skills were categorised as low difficulty (forehand-clear) and high difficulty (backhand-clear) according to the participation of the number of muscles, the co-ordination of muscles and joints and the experience of the athletes on them (Grice, 1996). Difficulty was operationally defined in terms of the level of technique combined with physical factors such as strength and power that would be required to perform the skill. Adding characteristics such as elevations, rotations, isolation and weight shift increased the difficulty level of a skill (Poon and Rodgers, 2000). Backhand clear is more difficult than the forehand-clear since the player does not have a vision of the playing court, and turns the body to the cross side of the target area so that aiming and depth perception is much more difficult. Also the follow through is not as big as the follow through of the forehand clear since the joints and muscles stop the movement and the athlete has to put much more power to direct the shuttlecock on the target area.

Procedures
All participants were pre-tested with the Poole and Nelson, 1993 badminton test as depicted in Figure 1 on the two badminton skills with different difficulty and their mean scores were not significantly different.

The participants were also tested on a self-confi-dence test (Vealey, 1986) for each skill. There was a pre test before the implementation of the feedback methods, an acquisition test 12 weeks after the implementation of the instructional methods and a retention test 2 weeks later. In the beginning of every training session a demonstration was performed by an elite player. The instructor answered any questions about the technique of the two badminton skills. The three experimental groups practised and received instruction 2 times a week, executing 10 exercises (Grice, 1996) on each skill and every exercise lasted 4 min. The instructor gave corrective feedback, approximately 10 times on each badminton skill, in every training session. Before the main part of the practice there was 10 min warming-up and 10 min rest period after the practice. During the training program there was a video camera placed at a distance of 6m and an ankle of 45o that recorded the performance of the athletes and was used to evaluate the outcome of the badminton skills tests.

The control group did not follow the practice. They participated only on the pre, post and retention tests. The three experimental groups followed the practice from the same coach but they were instructed according to the three different methods of corrective feedback. The first group (A) received positive feedback and instructional cues on how to correct the technique. The second group (B) received no positive feedback but instructional cues on errors of the execution. The third group (C) was a combined method and the participants received positive feedback and instructional cues on errors and directions on how to correct them. The instructors provided the participants summary feedback, every five trials, according to a written list of possible errors.

Badminton test
The purpose of the test (Poole and Nelson, 1993) was to measure the participants' performance ability executing the forehand-clear and the backhand-clear skills. The one side of the badminton court was separated into 4 parts (See Figure 1). An athlete was standing close to the last line of the opposite side of the court. He threw and hit the shuttle using the forehand-clear or the backhand-clear technique. Another athlete was standing on the other side in the middle of the half court (3.35m) holding the racket high (total high: 2.30m). The shuttle should pass over the net and the athlete's racket. The score depended on the point that the shuttle was landed. The athlete performed 12 shots, out of which the 10 best counted. The test retest reliability coefficient was satisfactory (α = 0.90).

Self confidence test
The State Sport Confidence Inventory - SSCI (Vealey, 1986) was used to evaluate self-confidence of the participants. The inventory was administered prior to each testing situation in order to assess the athletes' degree of confidence that they would be successful in that sport. Each item was measured on a 1 (low) to 9 (high) point Likert scale. The SSCI demonstrated good internal consistency, r = .95, and adequate concurrent validity, r = .64. The participants completed the questionnaire prior to the start of their testing.

Training of the instructor
The instructor was a certified coach. Prior to introducing the training programs, specific written directions were given to the instructor on how to implement the instructional methods. The motor program errors were correcting first and the parameter errors later (Magill, 1993) from a list of possible errors constructed according to the bibliography (Grice, 1996). A pilot test with another group before the test assured the understanding and the correct implementation of the methods by the instructor.

Statistical analysis
The independent variables were the different feedback groups (A, B, C and D), the levels of the skills difficulty (low and high) and the measurement periods (pre, post and retention). The dependent variables were the outcome and self confidence scores. Two three-way 4(Groups) X 2 (Difficulty) X 3 (Measures) analysis of variance with repeated measures on the last factor was applied, for the outcome and self confidence scores of each skill. Whenever a significant F ratio obtained a Scheffé post hoc test with a Bonferroni adjustment was applied to assess multiple comparisons of mean differences. Finally a regression analysis was conducted to determine the prediction of self confidence by the outcome. The data were analysed using the S.P.S.S. (15 ed.) package.

Effect measurement periods
Additional to the combined group (C) that was hypothesized, the positive feedback group improved outcome scores but not group (B). There was a significant main effect for the outcome scores for the low (F_2,45 = 3.685, p < 0.05) and the high difficulty level skill (F_2,45 = 41.969, p < 0.05) across the measurement periods. The Scheffé post hoc analysis revealed that all groups improved mean scores from pre to post and retention test for both skills except group (B) that received only error cues for the difficult skill or the control group.

Interaction of groups and difficulty level
The improvement of the outcome scores was different for the groups and the skill level and a significant interaction (F_4,90 = 13.04, p < 0.05) was found among the outcome scores for different measurement times and groups for the different difficulty skills (summary of the result in Table 2).

Self confidence
Effect Group
According to what was hypothesized, providing both positive feedback and error cues (group C) improved more the self confidence scores for both the easy and the difficult skills. However, it was found that group (A) improved also the self confidence scores for the easy skill. There was a significant main effect (F_2,45 = 27.323, p < 0.05) among the retention scores of the groups. A Scheffé post hoc analysis at retention for the easy skill showed that group (A) receiving positive feedback and instructions on correct execution of performance and group (C) receiving the combined feedback information were significantly better than group (B) that receiving error cues or the control group (D). Contrary, for the difficult skill at retention, only the group (C) receiving both positive feedback, correction and error cues was significantly better than the other three groups.

Effect measurement periods
According to what was hypothesized, the combined group (C) improved the self confidence scores across time for both the easy and the difficult skill. Additionally, the positive feedback group (A) improved outcome scores across time only for the easy skill. Finally, group (B) did not improved outcome scores for any skill across time. There was a significant main effect for the self confidence scores for the low (F_2,45 = 57.492, p < 0.05) and the high difficulty level skill (F_2,45 = 25.359, p < 0.05) across the measurement periods. The Scheffé post hoc revealed that group (C) that receiving the combined feedback, correction and error cues, improved mean scores significantly from pre to post and retention test for both skills, group (A) that receiving positive feedback improved mean scores significantly through time only for the easy skill and the group (C) that receiving error cues or the control group (D) did not improved mean scores significantly in any skill.

Interaction of groups and difficulty level
The improvement of the self confidence scores was different for the groups and the skill level and a significant interaction (F_4,90 = 11.78, p < 0.05) was found among the self confidence means for different measurement times and groups for the different difficulty skills (summary of the result in Table 3).

Relationship between outcome and self confidence
A regression analysis was conducted to investigate the proportion of the prediction of the participants' self confidence by their outcome scores. The outcome scores were entered as the independent variable and the self confidence scores as the depended variable. The outcome predicted a significant (F_(1,46) = 4.016, p < 0.01) proportion of the variance (16%) in the dependent variable for the easy skill and a significant (F_(1,46) = 21.559, p < 0.01) proportion of the variance (30.4%) for the difficult skill.

• The type of the skill is a critical factor in determining the effectiveness of the feedback types.
• Different instructional methods of corrective feedback could have beneficial effects in the outcome and self-confidence of young athletes
• Instructions focusing on the correct cues or errors increase performance of easy skills.
• Positive feedback or correction cues increase self-confidence of easy skills but only the combination of error and correction cues increase self confidence and outcome scores of difficult skills.