Helping Children Think: Gaze Aversion and Teaching.

Looking away from an interlocutor’s face during demanding cognitive activity can help adults answer challenging arithmetic and verbal-reasoning questions Schroeder, 1998). However, such ‘gaze aversion’ (GA) is poorly applied by 5-year old school children Bruce, Bonner, Longbotham, & Doyle, 2002). In Experiment 1 we trained ten 5-year old children to use GA whilst thinking about answers to questions. This trained group performed significantly better on challenging questions compared to ten controls given no GA training. In Experiment 2 we found significant and monotonic age-related increments in spontaneous use of GA across three cohorts of ten 5-year old school children ( M ages: 5;02, 5;06 and 5;08). Teaching and encouraging GA during challenging cognitive activity promises to be invaluable in promoting learning, particularly during early primary years. mid year end year type by cohort, F < 1; question difficulty by cohort, 27) = 1.96, η p2 = .13, p = .16; question type by question difficulty, F < 1; question type by question difficulty by cohort, F < 1.

Whilst GA occurs very little when people are listening to another person speak (e.g., Argyle & Cook, 1976;Doherty-Sneddon et al., 2002;Glenberg et al., 1998), it predominantly occurs whilst thinking and (albeit to a lesser extent) whilst speaking.
So, the occurrence of GA potentially reflects the need to concentrate on drawing information from memory and/or engage in on-line cognitive processing, such as speech-planning and mental arithmetic (Doherty-Sneddon et al., 2002;Glenberg, 1997;Glenberg et al., 1998). Conversely, given that, under normal circumstances, speech perception may be facilitated by the processing of visual information from a speakers face (e.g., Erber, 1979;Neely, 1956;McGurk & MacDonald, 1976), having access to relevant visual cues is likely to be most beneficial whilst listening to a speaker. In other words, we attend to visual cues when they are most useful to us, but when we need to concentrate on our internal cognitive processing we 'ignore' them by averting our gaze away from the person with whom we are interacting (Doherty-Sneddon et al., 2002;Glenberg, 1997;Glenberg et al., 1998).
Recent results have shown that this tendency to look away from an interlocutor's face increases when thinking about the solutions to increasingly difficult questions (Doherty-Sneddon et al., 2002;Glenberg et al., 1998; see also Doherty-Sneddon et al., 2000). For example, Glenberg et al. (1998;Experiments 1-3) found that adults' use of GA during the thinking stage of a question-answer interaction increased as arithmetic and verbal questions became more difficult. Eight year-old children have also been shown to increase the frequency with which they use Helping Children Think 4 GA during the thinking stage of a question-answer interaction as arithmetic and verbal questions prove increasingly difficult (Doherty-Sneddon et al. 2002). Given this finding, GA might be argued to be an overt sign of cognitive activity. Indeed, others have suggested that GA operates by enabling the thinker to concentrate and exert control over their own cognitive processing (see Doherty-Sneddon et al., 2001;Doherty-Sneddon et al., 2002;Glenberg, 1997;Glenberg et al., 1998). Consistent with this interpretation is the finding that GA occurs in response to objects other than faces, including video cameras (e.g., Kocel, Galin, Ornstein, & Merrin, 1972;Ehrlichman, Weiner, & Baker, 1974;Meskin & Singer, 1974). It appears then that people don't just avert their eyes from faces when they are thinking (see Ehrlichman, 1981), but also from any potentially distracting stimulus.
Of specific interest here, it appears that, under certain task conditions, engaging in GA whilst thinking can significantly benefit performance (e.g., Doherty-Sneddon et al., 2001;Glenberg et al., 1998: Experiment 4). For example, Glenberg et al. (1998: Experiment 4) required adults to answer easy, moderately difficult, and difficult arithmetic and general knowledge questions whilst either looking at an interviewer's face or closing their eyes. They found that when adults closed their eyes whilst thinking about moderately difficult problems they answered with greater accuracy as compared to conditions where they looked at the interviewer's face. No performance benefit was found for very easy or very difficult problems.
So, when questions prove challenging yet solvable, looking away from an interlocutor's face can benefit performance. This benefit of diverting one's attention away from potentially distracting visual cues has also been reported for children's performance on visuospatial memory tasks. For instance, when 6-and 10-year old children completed The Mr Peanuts Task (De Ribaupierre & Bailleux, 1994) and The Corsi Block Task (Corsi, 1972), their performance was poorer when they had to look at a face during a 10 second retention interval as compared with looking at the floor or at a moving visuospatial pattern (Doherty-Sneddon et al., 2001). Further, instructing 10-year old children to look at a face whilst forming a mental image of a verballydescribed abstract shape also impaired performance, as evidenced by both a decrease in accuracy and an increase in time to respond (Doherty-Sneddon et al., 2001; see also Doherty-Sneddon, McAuley, Bruce, Langton, Blokland, & Anderson, 2000). So, when we have to concentrate on internal processing of information, unhelpful or taskirrelevant visual cues can prove distracting, and consequently impair cognitive performance.
This finding of a functional benefit of GA during cognitively demanding activity is consequential given recent reports that GA is used inefficiently by certain populations. For instance, Doherty-Sneddon et al. (2002) have reported age-related differences in the effective use of GA: whilst GA is spontaneously employed by 5year old children, they tend to use GA less often and less consistently than both older (8 year-old) children (see Doherty-Sneddon et al., 2002) and adults (see Glenberg et al., 1998). So, compared to 8-year old children, 5-year old children have been shown to use less GA whilst thinking about and speaking their responses to both verbal and arithmetic questions (Doherty-Sneddon et al., 2002). Furthermore, whilst both 8-year old children (Doherty-Sneddon et al., 2002) and adults (Glenberg et al., 1998) have been shown to consistently increase their use of GA in response to increasing cognitive difficulty, 5-year old children have been reported to do so only inconsistently (see Doherty-Sneddon et al., 2002).
Since it is difficult to draw any causal links between GA and performance benefits in children from this earlier work of Doherty-Sneddon et al. (2002), the main 6 motivation in the current study was to see whether this is the case. We selected 5-year old children following Doherty-Sneddon et al.'s (2002) earlier report that such young children use GA inconsistently and relatively infrequently; an important finding given the reports that GA can benefit performance on a range of tasks requiring cognitive processing (e.g., Glenberg et al., 1998;Doherty-Sneddon et al., 2000;2001). Fiveyear old children therefore represent a group for whom there is strong potential for increasing spontaneous levels of GA, thus enabling an ideal opportunity to measure any benefits of increased use of GA on task performance. Indeed, if any functional benefits of GA training can be shown with 5-year old children, this would provide an educationally relevant strategy for teaching young children.
The purpose of the experiments reported here is to examine the role of GA in 5-year old children, focussing on the extent to which they may be encouraged to adopt GA as a behaviour (Experiment 1), the extent to which its use may facilitate performance (Experiment 1) and the extent to which spontaneous engagement in GA develops throughout the first year of formal education (Experiment 2).

Experiment 1
In Experiment 1, we wanted to see whether we could train a group of 5-year old children to increase the proportion of time they spent looking away from a questioner's face whilst thinking about answers to arithmetic and verbal-reasoning questions. In addition, we wanted to see whether any increase in GA for 5-year old children would lead to an improvement in response accuracy.

Participants
Twenty 5-year old children were recruited from an LEA primary school in Stirlingshire. Ten were randomly allocated to a control condition (6 boys, 4 girls, M 7 age = 5 years 2 months, range = 4 years 9 months to 5 years 4 months) and ten to an experimental (avert group) condition (6 boys, 4 girls, M age = 5 years 1 month, range = 4 years 9 month to 5 years 5 months). Planned comparisons showed groups were comparable in terms of age, t(18) = 0.53, p = .61.
Examples of the type of stimuli used are included in Table 1.

Procedure
Children were asked the questions individually in a quiet location separate from their classroom, and were seated directly facing the questioner at a distance of approximately 1.5 feet, thus enabling the questioner to maintain her gaze on the child's face throughout testing. The questioners' locus of gaze (i.e., the child's face) was held constant in both conditions since the critical manipulation of interest was to see whether 5-year old children could be successfully trained to increase their use of GA during one-to-one pedagogical interactions, and to establish whether this trained behaviour would significantly benefit performance. Prior to test proper all children first answered 20 practice questions. At this stage, children in the avert group were instructed as follows: 'I'm going to ask you some questions. Once I finish asking the question I want you to look away from my face and try to think of the answer. If you don't know the answer it's okay -just so long as you try your best.' Children in the control group were given similar instructions: 'I'm going to ask you some questions.
Once I finish asking the question I want you to try to think of the answer. If you don't know the answer it's okay -just so long as you try your best.' All children were asked the same set of practice questions once. Test proper followed the practice session immediately, lasting approximately 15 minutes per child. During test proper the avert group were instructed to continue using GA whilst thinking about their answers to questions, whilst the control were again given no instruction as to where to focus their gaze whilst thinking. Ordering of question type, question difficulty, and the individual questions was fully counterbalanced across participants. To enable quantification of children's direction of eye gaze during the thinking stage of the question-answer interaction, a front-on view of each child's head and shoulders was video-recorded throughout testing using a digital camcorder. The presence of the same experimenter and filming technique was maintained in both conditions to preclude the possibility that any benefits of GA training be attributable to either factor. Whilst all children were fully aware that they were being filmed, they were not aware that their gaze behaviour was of specific interest.

Results and Discussion
The effects of task difficulty and gaze training on both response accuracy and the percentage of time that gaze was averted during the thinking episode (i.e., the interval between the questioner having finished speaking and the child having started speaking) were analysed separately. For the accuracy data, correct answers were given a score of 1 and incorrect answers a score of 0. These data were then used to calculate the percentage of correct answers for each child under each condition. For the GA data, the proportion of time spent looking away from the questioner during the thinking episode was calculated to millisecond resolution. Interjudge reliability as to whether GA had occurred was calculated for a random sample of 10% of the participants. In total 88 episodes were coded by two judges, for which there was 92.05% interjudge agreement. Furthermore, the coders' scoring for the duration of GA correlated significantly, r (87) = .69, p < .0001.

Proportion of Thinking Episode Engaged in Gaze Aversion
A 2 (question type: arithmetic, verbal reasoning) X 2 (question difficulty: easy, moderately difficult) X 2 (group: control, avert) mixed design ANOVA was carried out on the proportion of time spent averting gaze whilst thinking. Question type and question difficulty were the within-groups variables, and group the betweengroups variable. Means for each condition are displayed in Table 2. INSERT The results of Experiment 1 have shown that it is possible to encourage 5-year old children to increase their use of GA when thinking about the answers to arithmetic and verbal reasoning questions. We found an increase in GA for the avert group relative to the control group for each type of question at each level of difficulty.
However, there was no significant increase in GA for the avert group relative to the control group for easy arithmetic questions. Whilst this finding occurs because controls used substantially higher levels of spontaneous GA overall for arithmetic questions than verbal questions, coupled with the fact that the avert group used least GA for easy arithmetic questions, we don't attach any significance to this observed interaction since it likely reflects a ceiling effect for easy arithmetic questions, as we discuss below. To summarise, these data clearly demonstrate that, given minimal training (practice of GA on a sample of 20 pre-test questions), 5-year old children can be encouraged to substantially increase the proportion of time they spend engaged in GA while thinking.
Further, our results show that 5-year old children's ability to answer arithmetic and verbal questions can be significantly improved as a result of GA training. This finding is novel, since there have been no previous reports of performance benefits for children answering arithmetic and verbal-reasoning questions under conditions which mimic classroom one-to-one teacher-child interactions. So, consistent with the adult literature (Glenberg et al., 1998), for arithmetic and verbal questions of moderate difficulty, we found a significant increase in response accuracy. Inconsistent with the adult literature, we further found a significant increase in response accuracy for easy verbal questions. Whilst there was no significant increase in response accuracy for the easy arithmetic questions, inspection of accuracy data for this condition suggests that controls were already performing at ceiling in this condition (see Table 3).
This pattern of results would suggest that any benefits of GA are limited to situations where there is room for improvement in performance. So, when questions prove trivial (as with our easy arithmetic questions), GA is unlikely to help the thinker solve the question. However, when questions prove challenging (as with our moderately difficult arithmetic and verbal questions and the easy verbal questions), GA can potentially enable the thinker to arrive at the solution to that question. This interpretation may also account for our finding that whilst the control group spontaneously increased their use of GA in response to question difficulty, they did so only for the verbal questions. In addition to these main findings, we also observed another interesting pattern of behaviour when we considered the spontaneous GA rates of the control children.
Using similar questions to those currently used in Experiment 1, Doherty-Sneddon et al. (2002)  study with that of our 'control group' from Experiment 1. Added to this, in order to provide a more detailed picture of the changes that may occur in GA during primary year 1, we also collected new data on the spontaneous use of GA for a group of children who were in the middle of primary year 1. All cohorts of children were asked verbal and arithmetic questions with two levels of difficulty (i.e., easy and hard).

Participants
Three cohorts of five-year old children were used. The first cohort comprised the 'control group' of Experiment 1, representing ten children who had just entered

Stimuli
As in Experiment 1, for each cohort, the questions used were drawn from the items described in the WPPSI, and were either verbal (cohort 1, n = 26; cohort 2, n = 24; cohort 3, n = 24) or arithmetic (cohort 1, n = 18; cohort 2, n = 24; cohort 3, n = 14). (See Table 1 for examples of stimuli used). Following the same criteria as in Experiment 1, consultation with the participants' teachers established that half of the questions were easy (predicted accuracy 80%-100%) and half difficult (predicted accuracy 30-50%).

Procedure
For each cohort, the procedure was the same as that of the 'control group' of Experiment 1.

Results and Discussion
The effects of task difficulty and cohort on percentage of time that gaze was averted during the thinking episode were analysed.

Spontaneous Gaze Aversion During Thinking stage
Analyses of spontaneous GA were conducted to see how its use changes over the child's fifth year/first year of formal education. A 2 (question type: arithmetic, verbal reasoning) X 2 (question difficulty: easy, difficult) X 3 (cohort: start year 1, mid year 1, end year 1) mixed design ANOVA was employed, with question type and question difficulty the within-groups variables, and cohort the between-groups variable. Means for each condition are displayed in Table 4.   INSERT TABLE 4  So, although GA is less frequently applied at the onset of primary year 1, it still acts as a reliable overt signal that a child is engaged in challenging cognitive activity.

General Discussion
In Experiment 1 we demonstrated the efficacy of instructing 5-year old children to increase their use of GA whilst thinking about their answers to arithmetic and verbal-reasoning questions. The performance benefits which arose as a result of this increased use of GA were limited to questions which proved challenging to the children, with no advantage for using GA while solving easy problems. This pattern of results mirrors that reported in the adult literature (Glenberg et al., 1998: Experiment 4), and concords with the assertion that GA is a behaviour which grants thinkers the opportunity to concentrate on internal processing of information (e.g., Doherty-Sneddon et al., 2002;Glenberg, 1997;Glenberg et al., 1998).
The finding that GA serves to facilitate performance when children (and indeed adults) are challenged could be taken to suggest that the best time at which to encourage a child to use GA would be when they are challenged, but working within their current cognitive understanding of a given operation, or in Vygotsky's (1934Vygotsky's ( /1962 terminology, when working within their 'zone of proximal development'. Indeed, preliminary work (Longbotham, 2001)  Given the observed benefit of GA, it is important to note a developmental trajectory in the use of this behaviour. So, whilst 5-year old children at the onset of Primary 1 were beginning to use GA to exert control over their cognitive processing, use of this behaviour had become more frequent by the time children were nearing the middle of their fifth year. Consequently, GA, at any age, can be considered an overt sign that a child is engaged in cognitive activity. Given the relatively small sample sizes used in the current experiments, a certain degree of caution should be exercised in considering the implications of these results. However, despite the small sample sizes a clear influence of GA on facilitating thought and concentration was observed, thus highlighting the importance of GA during pedagogical interactions.
These findings have potentially important implications for children's attainment and for teaching methods. First, GA should prove useful as a tool by which to identify children who are engaged in cognitive activity. More importantly, children actually benefited from appropriately-timed GA on tasks typical of everyday pedagogical interactions. So, GA whilst thinking is a simple behaviour by which children can more effectively exert control over their own cognitive processing, potentially offering a means to more effective learning. This is an important finding for teachers seeking to provide optimal support to children engaged in problem solving activities.

Footnotes
Removed to protect author anonymity.