Performance and fine motor control abilities

The CKAT scores did not significantly differ between groups, [F (2,60) = 1.365, p = 0.263, ηp2 = 0.044], or between male and female participants [ t (61) = 1.492, p = 0.141]. A Spearman's rank-order correlation was performed to assess the relationship between the overall performance scores and CKAT battery scores. There was no correlation between CKAT and errors [rs (61) = 0.128, p = 0.319].

6.7 Discussion

Novice participants were taught a basic manual dexterity task within a VR haptic simulator using qualitatively different types of feedback during training. We found that the participants who received a combination of instructor-led and VR haptic simulator feedback adopted a more cautious strategy than those who were exposed to one type of feedback alone. Specifically, they produced fewer errors and also removed less of the target than the other groups. We suggest that such behaviour is potentially advantageous for novice trainees - producing safer practice relative to an over ambitious student

sacrificing accuracy for greater target removal. Importantly, we also demonstrated that the presence of VR devices alone is not sufficient for optimal training of motor skills and must be coupled with expert guidance. Our findings are consistent with the motor learning and medical literature which indicates that multimodal feedback is more effective than unimodal feedback- particularly during the early acquisition of complex skills (Sigrist et al. 2013; Hatala et al. 2014). Whilst others have previously shown the value of providing augmented visual feedback with additional tuition sessions prior to training (Wierinck et al. 2006b), our work presents the first set of data

demonstrating the value of haptic simulator feedback combined with continuous instructor feedback in motor skill acquisition and retention. Although there were some differences among the three groups in baseline (pre-test) performance, this trend was not statistically significant (in any of the performance measures), and did not explain performance differences at subsequent experimental stages.

The device feedback group exhibited significantly higher task completion scores in less time compared to the other two groups, however, this has impacted the accuracy of performance as they also have higher error scores and subsequently lower overall performance. Since there was no time limit in any experimental phase, this trend could potentially be explained by the presentation mode of the performance measures on the computer screen, so their attention was directed toward the first information appeared on screen (which was task completion score), with little or no attention given to other measures such as error scores, because there were no guidance provided during training phase as what important measures to look for when they want to evaluate their progress. The finding that the group who received feedback from the device alone was the lowest performing throughout the experiment is instructive for the teaching of motor control skills in dentistry. Research on motor skill acquisition indicates the existence of two broad mechanisms that interact and contribute to learning any given motor task (Haith & Krakauer 2013). The most rapid method of improving task performance is known as “model-based” (MB) learning and depends upon previously developed ‘forward models’ that allow the trainee to make predictions about the consequences of their actions. This is the type of mechanism that most likely underlies the process of learning to use dental loupes (i.e. where an experienced dentist will use existing knowledge about task-related perceptual information to calibrate to a new visual environment in order to perform a task). Although MB learning is initially a cognitively expensive activity, the speed of skilled acquisition can lead to relative automaticity of performance in a short period of time (Haith & Krakauer 2013). The second form of learning is known as “model-free” (MF). This learning involves the development of ‘inverse models’ or ‘controllers’ via

trial and error learning and is a slower process. MF learning is an essential component of skill acquisition and would underpin the learning process within all three of our experimental groups. But the provision of additional information allows individuals to exploit MB learning processes and generalise their skills to situations that have not been previously encountered. In line with this framework for understanding motor learning, the present data suggest that excessive error can be reduced through guidance from an external source such as an experienced instructor (i.e. the IDFB group). This guidance provides information that can be used rapidly to develop forward models specifying appropriate task-related actions. Evidence that participants in the IDFB group were able to achieve such a feat is demonstrated by the finding that their skill levels were consolidated over time and that information learnt in one task could be generalised to another, thus demonstrating rapid near transfer (Schmidt & Wrisberg 2008) - a hallmark of MB processes.

The aim of the transfer tests is to evaluate skill learning using similar task at different training settings or different task and same training condition. The transfer phase in the present study comprises two tests, both using different tasks (abstract shapes) that were not encountered during training. This

particular type of transfer is referred to as near transfer or generalization of the learned skill (Schmidt & Wrisberg 2008). Typical transfer test would be

performed in different training environment (such as phantom head simulator). Our results showed that all groups exhibited lower overall performance scores at the transfer phase, particularly a sharp rise in error A and error B scores (Figure 6-8), although not statistically significant. This could be attributed to the unfamiliar features of the new shape that is not encountered during training,

which is a transition from straight outline to circular and cross abstract shapes that need a slightly different preparation approach.

In the current study, the concurrent verbal feedback from the instructor is a form of knowledge of performance (KP) that did not simply indicate the presence or absence of errors, but also elaborated on how to enhance performance by directing the trainee attention to important performance aspects. Additionally, the participants were encouraged to enquire about any performance related information, therefore moving away from unidirectional instruction to dialogue-rich learning context.

In the IFB group, receiving verbal feedback alone without any visual display of performance measures from the simulator allow the student to depend solely on the verbal comments and instructions from the tutor to guide the

performance. This represents a cognitive effort to understand and process what should be done until reaching a mental representation of the desired outcome before actually executing the action. This was evident in the overall trend of IFB group to spend longer time performing the task throughout the experimental phases (Figure 6-8), although this was not statistically significant except at transfer test. Additionally, the student may not be able to know what exactly contributed to good or otherwise unsuccessful performance (i.e. whether it is the leeway sides or the bottom error scores that was more serious).

On the other hand, the superior performance of IDFB group could be related to the instructor directing the trainee attention to task relevant information and specific automatic measures that are generated by the simulator, ultimately combining the important aspects of the learned task from both sources and

minimizing the distraction from redundant or irrelevant information (Kalyuga 2008) which prevent cognitive overload and enhance performance. The flow of information received by the student is processed more efficiently when it is distributed over multiple sensory modalities, because of the specificity of the human senses that process various types of information differently (Sigrist et al. 2015).

It is worth noting that whilst reducing error through instructor feedback was useful for our sample of novice trainees, error augmentation could provide a more effective means of accelerating learning in a group with a higher level of skill (Chen 2001). In other words, the amount of assistance and pedagogical feedback provided to final year undergraduates to achieve mastery of a task is likely to be qualitatively different to the optimal strategy for trainees earlier in their training. Task difficulty is also likely to modulate the relationship between optimal feedback and motor learning. For example, the optimal feedback for a basic manual dexterity exercise might be different to that required for a Class II cavity preparation or during the application of restorative materials. This could be further explained on the basis of what abilities constitute the main skill ( see Table 2.3-a), for example in manual dexterity training, feedback is focused on how to hold the handpiece correctly, on hand-eye coordination, cutting

pressure and direction etc., while training on more specific procedures such as Class II cavity preparation will require a more precise feedback about a

smooth conservative preparation, retention and resistance forms and avoiding damage to the adjacent teeth etc. It follows that the type of feedback provided during preclinical and clinical dental training needs to be carefully considered and investigated in order to ensure optimal learning. In fact, no feedback is

sometimes necessary as in the case where the students are experienced and highly motivated to monitor their own performance (Chen 2001).

The introduction of VR simulators with the real-time quantitative evaluation of the student input, add another dimension into how feedback is delivered to dental students. Based on these objective simulator-generated metrics further assessment can be provided such as objective augmented feedback to

students, errors detection and correction, as well as informed decisions about student’s competence in particular procedures (Porte et al. 2007).

It is critical that dental educators investigate the best pedagogical approaches to utilize such kinematic information to enrich the feedback practices during the simulation experience. They need to draw on key theoretical concepts from other disciplines (particularly education literature, motor skill learning literature) and adapt the best practices to suit our specific needs and ultimately optimize the dental student learning experience.

6.8 Conclusions

The learning of basic manual dexterity skills was accelerated when participants were provided with haptic simulator’s feedback in conjunction with an

experienced dental instructor feedback, relative to groups with access to the device only or instructor only feedback. This was particularly beneficial for the retention of learned skills. There was an overall performance improvement for all groups at the end of the experiment (retention phase), which was evidenced by lower error scores as well as comparable time for task performance (DT). These findings were supported by evidence from motor learning literature which describe the rapid “model-based” learning, based on previously developed ‘forward models’ (as seen in the combined feedback group

performance), and the slower “model-free” learning based on ‘inverse models’ via trial and error (as seen in the device feedback group performance).

These data indicate that integration of VR into a dental curriculum needs consideration in order to maximise VR's potential utility in motor skill learning and to complement existing simulation techniques. This will be discussed further in the next chapter.

Chapter 7 : Integrating Virtual Reality Haptic simulation into