4.2 The Comparison of Behavioural Measures across Studies revealed consistent
4.2.3 Timing Differences are the Most Probable Cause of the RT Effects Measured
The pattern of the response types of the fMRI study met the previously discussed points of the pilot and MEG studies: The increase in HR and RR of match responses and an increase in discriminability indicated a positive learning progress referring to familiarization acquisition. The discriminability increase measured by the fMRI study confirmed the refinement of the internal model across the sequential positions by the same pattern: An increase was observed from the first to the third sequential positions. The missing response bias effect in the fMRI study diminished the hazard of the discriminability impeded by response bias. Further, the fMRI study confirmed the highest contrast for discriminability of the MEG study with the second simple effect, which was relevant for neurophysiological investigations. Hereby, we conclude the three studies showed converging evidence.
Remarkably, for the fMRI study a significant RT decrease was observed. The decrease was observed for RT for HR and RR for the match responses. All this observation refers to for the second simple effect. Thus, this observation is in line with our initial hypothesis that the learning should be measured with an increase in HR and a decrease in RT. According to our assumptions, we did not observe any significant effect regarding the non-match condition. However, we analysed the RT for the sequential position, regardless of the response type, and also observed a significant decrease in the second simple effect. The non-match responses also contribute to this effect, and, thus, we cannot completely rule out that an increase in the RT is solely due to the learning and that a certain improvement in the task per se drives the RT decrease as well in form of a contributing factor.
However, we argued earlier that the missing RT effect in the pilot study (and MEG study) was due to task instructions, but the task instruction did not change between these two studies. The question arises as to why do we measure significant RT decreases with the fMRI paradigm but not in the pilot and MEG paradigms. The factor which can be objectively examined are the alteration of the timing parameters in the experimental setting. Whereas the timing parameters in the pilot and MEG studies did not vary in the delay and cue latency, both were jittered in the fMRI. The jittered delay between sample and test stimulus led to a Stimulus onset asynchrony (SOA) of the test stimulus. We propose the best explanation why RT effects were observed arise with the SOA.
Support for this is that the use of a high SOA in fMRI studies is been done to separate the haemodynamic response function (HRF) to the corresponding events, thus, the conditions are able to separate for statistical analysis (see SPM guidelines). According to that, this means the stable delay between sample and test stimulus led to entangled sample and test post-stimulus processes in the MEG (and Pilot) study which is disentangled in the fMRI study. Therefore, applying SOA the proportion of RT decrease due to learning across the sequential positions becomes measurable in the fMRI experiment. Consequently, for the MEG (and Pilot) study the delay led to carry-over post- stimulus processes of the sample stimulus into post-stimulus interval of the test stimulus. However, the carry-over of processes does not alone explain the missing RT effect in the MEG (and Pilot) studies. In the assumption of the carry-over effect, another consequence is that the transferred process must counteract the RT-related learning effect, otherwise
RT effects would have been measured in all studies. In the following, we provide an explanation why sample post-stimulus processes most probably become longer.
We assume that across the sequential positions the PE become more which might be the contributing factor why sample post-stimulus processes increase. Again, we explain this with reference to the processing model. The early internal model is raw and contains less details, the consequence of this is that the early matching processes can be associated with rather raw update processes then detailed evaluation processes of the internal model. However, the later becomes the case with increasing details of the internal model. Thus, the PE become more detailed, but also become another quality due to an intermediate internal model which contains more and more sufficient information. This information is sufficient enough to separate the visual information input into new visual information and known information and further to separate the new information in conflicting (when features of a false positive are integrated) or coherent new visual information. Thus, these PE contain a different quality as that they led to reject information of the intermediate internal model to become a mature internal model. This concept of PR driving model refinement and referring this mechanism to familiarity acquisition was first described by Apps and Tsakiris (Apps and Tsakiris, 2013). Consequently, the probabilities of PE and related processes increased across the sequential positions, and, therefore the sample post-stimulus processes increased. This can be one explanation why RT effect were not measured in the MEG and Pilot experiment.
In summary, the measured RT effect in the fMRI experiment did meet all our hypotheses. We observed a positive learning effect by an increase in HR and a decrease in RT. The fMRI study revealed explanations for the missing RT effects in the MEG study by the presence of a SOA. We proposed that sample post-stimulus processes increased across the sequential positions due to PE related processes. Further, we proposed that these processes carry-over into the test stimulus interval and here counteracts the RT increase due to learning.