Experiment 3b Move & Fixate Training in concert: The affect of task

The methodology of the final experiment of this chapter is identical to last study described: the search-array paradigm was used, except NFC stimuli were presented. Therefore the only difference is task; training was operationalised in the same way as before, with four groups, NT, FT, MT, and BT. When untrained in block 1 (or throughout for the NT group) the task required participants to first locate a uniquely defined target on the basis of its number of shapes (sub-arrays of 3 or 5 shapes being targets), then to respond to it on the basis of its feature conjunction (with one button press designated for red circles and green squares, and a different button press designated for green circles and red squares) as in Experiment 1. Move training advised that the location of the target in the grid was predictable from trial- to-trial, as prescribed by the egg-timer pattern; Fixate training advised that the correct response to red circles and green squares always held to the rule that the stimulus’ central region would contain a shape, while this was never the case for green circle or red square targets; with combined Move and Fixate training participants were informed of both of these contingencies.

As the stimulus discrimination judgement on responding pertains to feature conjunctions in Experiment 3b, and such feature integrations should be very

expected which can be credited to Fixate training, either independently (with the FT group) or supplementary to Move training (with the BT group).

Also, given the argument outlined when discussing the previous experiment (section 2.3.3), if the target identification aspects of both the FCN and NFC tasks are equated in the visual search grid, and the greater difficulty of processing feature conjunctions relates only to the stimulus discrimination judgement on responding, then one should observe slower reaction times in block 1 with Experiment 3b compared to Experiment 3a. Moreover the relative changes in RT’s following Move or Fixate training should reflect both the search and response aspects of the task separately. Therefore it is possible to estimate any changes in these components between experiments, so if the change in task affects visual search or stimulus discrimination it should be possible to discern this.

The main hypothesis therefore concerns additive training advantages which should be evident owing to the change of Task in the present experiment. However, a secondary prediction is linked to the first hypothesis: If responding according to feature conjunction is more visually and cognitively challenging then it is possible that this will induce competition between the Move and Fixate centres at the end of each trial when participants are directed with MT in isolation. The ‘push-pull’ mechanism for maintaining equilibrium between the Move and Fixate centres (Findlay & Walker, 1999) may express this competition in terms of an impediment with the response discrimination (reflected by a decrease in accuracy); however it is also conceivable that a diminished ability to execute the MT scanpath will be observed (reflected by a smaller RT decrease for the MT group than we have seen in the last two experiments). Either way, Fixate training should abolish any negative consequences associated with training the Move centre independently.

2.4.1. Method

As the only difference between the methods of this experiment and those of Experiment 3a was the use of NFC stimuli as opposed to FCN, specific details are not repeated again here. The task is described above, and new participants were recruited to perform it as detailed below.

Participants

Forty naïve paid participants (27 female) were recruited from the University of Nottingham’s student population and surrounding local area (mean age 25yrs, range 19-33). All participants reported normal or corrected-to-normal vision. The sample size of each training group was uniform (N = 10), and assignment was random as previously.

Four participants were replaced (one female from the NT group, one male and one female from the FT group, and one male from the MT group) because they did not complete the task properly; with accuracies

their group mean. These participants were substituted with four females aged 32, 24, 21 and 30yrs.

2.4.2. Results

Reaction Times

Reaction Time (RT) data were extracted from the data set for each of the two experimental blocks across all four training conditions. Again a check was carried out for anticipatory responses (<200ms), so that one could be sure that responses made were to the present trial. A tiny amount of trials fell into this category (0.39%) and

thus were removed from all subsequent analyses. Only correct responses were included in the RT analysis (25% of responses were incorrect).

Reaction time data were analysed the same way as in Experiment 3a, using a 2x4 mixed factorial ANOVA with two levels of Block, and four levels of Training. A significant main effect of block was observed (F (1,36) = 869.8, MSE = 19996.0, p <0.001), reflecting the lower mean RT in block 2 (x = 1758ms) compared to block 1 (

x = 2691ms). Further, there was a significant main effect of training (F (3,36) = 55.4,

MSE = 82558.1 p <0.001), evident in terms of the MT and BT groups (x = 1816ms; x

= 1806ms, respectively) performing considerably faster than the NT and FT (x ₌

2647ms; x= 2631ms, respectively) overall. Crucially however there was a significant interaction between block and training (F (3,36) = 163.51, MSE = 19995.95, p <0.001), which reveals differential performance at block 2 depending on the type of training applied (see Fig. 2.13).

500 1000 1500 2000 2500 3000 1 2 R e a c ti o n T im e s Block No training Fixate training Move training Both training

Fig. 2.13. Reaction times (ms) in block 1 and 2 for each training group. Error bars represent standard error of the mean.

The relative effectiveness of each training strategy was assessed as before with the difference score analysis. The one-way ANOVA with training as the grouping variable was significant (F (3,36) =163.5, MSE = 39991.89, p <0.001). This indicates training improved performance to different degrees. Post-hoc comparisons (with Tukey’s HSD) confirmed that the improvement between groups was significantly different in all cases at p <0.05 (the MT vs. BT comparison was significant at p <0.01, and the comparisons between groups which used Move training, either

independently or in concert with Fixate Training, and those which did not were significant at level p <0.001). This shows that with the NFC task, Fixate training, either alone or together with Move training, is effective in reducing stimulus processing time associated with the response discrimination judgement (see Fig. 2.14).

To test for comparable competence with the task on commencement of the experiment pre-training block 1 performance was checked with a one-way ANOVA. No statistically reliable differences were found (p >0.05), which is encouraging.

It was mentioned when introducing this experiment that if the prediction that NFC stimuli are harder in the grid array paradigm proves correct then, not only

0 200 400 600 800 1000 1200 1400 1600 1800 2000 NT FT MT BT R e d u ct io n in R e a ct io n T im e s Training Group

Fig. 2.14. Reaction time decrease (ms) from block 1 to block 2 for each training group. Errors bars represent standard error of the mean.

should a Fixate training benefit be observed, as we have seen, but also RT’s relating to general performance without training should be longer in experiment 3b

compared to experiment 3a. An independent samples t-test compared the overall mean block 1 performance in Experiment 3a to that of the current experiment. This analysis confirms the prediction (t99 = -9.7, p <0.001): NFC stimuli give rise to longer overall RT’s without training (Exp 3a: x= 2205ms vs. Exp 3b: x= 2691ms), likely reflecting their greater general difficulty manifest in the response discrimination to feature conjunctions. This would explain the increased susceptibility to Fixate training in the present experiment.

However, because of the difference between Experiments 3a and 3b in the effectiveness of Fixate training – the advantage offered being small in both cases, and only statistically significant in the latter – one final analysis is required to confirm that Fixate training offers a genuine advantage with the NFC, but not the FCN, sub- array task. Two between-experiments comparisons were conducted, the first analysing the absolute RT scores, the second analysing the difference scores for this measure.

The same pattern of results already reported can be confirmed with reference to these statistics. First, the general linear model ANOVA with two between-groups factors (Experiment, with two levels: Exp 3a and Exp3b; and Training, with four levels: NT, FT, MT, BT) and one within-groups factor (Block, with two levels: block 1 and block 2), produced significantly different results for all comparisons (lowest F = 7.7; highest F = 1388.3; highest p = 0.006). The most

important result here is that Experiment x Training x Block three-way interaction was statistically significant, as this supports the finding of differential Training efficacy depending on task. The second, univariate analysis (difference scores compared with Experiment and Training as between groups factors) mirrors these effects in the

omnibus measure (all comparisons showing differences: lowest F = 8.0; highest F = 184.9; highest p = 0.006). Moreover, Tukey HSD post-hoc contrasts comparing the training groups [collapsed across experiments] revealed significant differences in every case. Crucially, the NT and FT groups differed (p =0.044), and the MT and BT groups differed (p =0.001) (the remaining comparisons were significant at p <0.001). Given that analyses confined to each of the two experiments separately did not reveal an effect of Fixate training in Experiment 3a but did in 3b, it is quite parsimonious to conclude that Fixate training has greater utility in Experiment 3b than 3a, and this is what drives the effects of Fixate training when the results of these two experiments are included in an overall analysis.

Accuracy

Contrary to Experiment 3a, when the accuracy analysis is repeated for Experiment 3b, the task is now hard enough for a training advantage to be present. A main effect of block was again found, showing the general improvement associated with practice or training (F (1,36) = 232.9, MSE = 35.1, p<0.001), plus a main effect of training (F (3,36) = 9.5, MSE = 201.0, p<0.001), highlighting the overall difference in performance percentages between groups. The NT and FT groups were markedly less accurate (x = 66.6%; x = 66.0%, respectively) than the MT and BT groups (x = 83.6%; x = 82.7%, respectively) overall. However, these effects present as a consequence of a block * training interaction (F (3,36) = 6.82, MSE = 35.1, p=0.001); with participants responding more accurately in block 2 depending on the training strategy(ies) they employ (see Table 2.4). (It was also noted that upon debrief no participant reported guessing any of the training contingencies).

Block 1 Block 2