Trials 2 Second Edition 1.06 Serial
Note that our analyses focused on the two extreme cases of repetition trials with one versus eight repetitions of the first or second item while the experiment also included repetition trials with intermediate levels of repetitions (see SI). Specifically, other repetition trials included cases in which the second item began to appear at each possible position from 2 to 9. The other repetition trials could therefore include, for instance, three repetitions of the first and six repetitions of the second image, or four repetitions of the first and five repetitions of the second item, etc. The results reported in the SI indicate that effects in these trials show smooth transition between the extremes shown in the main manuscript.
trials 2 second edition 1.06 serial
In order to analyze the neural activation patterns following the presentation of sequential visual stimuli for evidence of sequentiality, we first determined the true serial position of each decoded event for each trial. Specifically, applying the trained classifiers to each volume of the sequence trials yielded a series of predicted event labels and corresponding classification probabilities that were assigned their sequential position within the true sequence that was shown to participants on the corresponding trial.
We hypothesized that sequential order information of fast neural events will translate into order structure in the fMRI signal and successively decoded events in turn. Therefore, we analyzed the fMRI data from sequence trials for evidence of sequentiality across consecutive measurements. The analyses were restricted to the expected forward and backward periods which were adjusted depending on the sequence speed. For each TR, we obtained the image with the most likely fMRI signal pattern based on the classification probabilities. First, we asked if we are more likely to decode earlier serial events earlier and later serial events later in the decoding time window of 13 TRs. To this end, we averaged the serial position of the most likely event at every TR, separately for each trial and participant, resulting in a time course of average serial event position across the decoding time window (Fig. 3d). We then compared the average serial event position against the mean serial position (position 3) as a baseline across participants at every time point in the forward and backward period using a series of two-sided one-sample t-tests, adjusted for 38 multiple comparisons (across all five speed conditions and TRs in the forward and backward period) by controlling the FDR133. These results are reported in the SI. Next, in order to assess if the average serial position differed between the forward and backward period for the five different speed conditions, we conducted a linear mixed effects (LME) model and entered the speed condition (with five levels) and trial period (forward versus backward) as fixed effects including by-participant random intercepts and slopes. Finally, we conducted a series of two-sided one-sample t-tests to assess whether the mean serial position in the forward and backward periods differed from the expected mean serial position (baseline of 3) for every speed condition (all p values adjusted for 10 comparisons using FDR-correction133).
The best account of benchmark order phenomenon in the verbal domain comes from modelsrelying on positional codes to represent serial order information (see, for example,Brown et al., 2000; Burgess & Hitch, 1999;Hartley et al., 2016;Henson, 1998; Lewandowsky & Farrell,2008). In positional models, serial order is represented by associationsbetween items and independent markers representing positions. The main strength of thisclass of models is its ability to account for temporal grouping effects. Temporalgrouping is characterised by the insertion of additional pauses between some itemsduring sequence presentation, inducing the perception of temporally distinct sub-groupsof items. With verbal material, such manipulations lead to the well-replicated phenomenathat constraint serial order models of STM (see Frankish, 1985, 1989; Hartley et al., 2016; Henson, 1996; Hitch et al., 1996; Maybery et al., 2002; Ng & Maybery, 2002, 2005; Ryan, 1969a, 1969b). For grouped sequences, a recalladvantage as well as a multiply-bowed shape serial position curve are usually observed.An increase in the proportion of interposition errors, or between-group displacements ofitems that keep their initial within-group serial position, is also characteristic ofthe recall of grouped sequences. For instance, in a 6-item sequence composed of twogroups of three items, an interposition error would be to recall the item from Position2 (i.e., Position 2 in the first group) at Position 5 (Position 2 in the secondgroup).
The experiment was separated into two blocks. In the first block,participants were presented with ungrouped and grouped sequences in thesecond block. The order of the trials presenting phonologically similar anddissimilar letters was random. Each block started with four trials,presenting two phonologically similar and dissimilar letter sequences.During the training session, participants had feedback regarding theaccuracy of their response, but not during the experimental trials.
Top panels: serial position curve; middle panels: transpositiongradients: bottom panels: response latencies. Left and right partsof the figure depict data from phonologically dissimilar and similartrials, respectively (Experiment 2). Error bars represent confidenceinterval computed on data corrected for between-subject variability(Morey,2008), following Baguley (2012, formula 8)recommendations.
Top panels: serial position curve; middle panels: transpositiongradients: bottom panels: response latencies. Left and right partsof the figure depict data from phonologically dissimilar and similartrials, respectively (Experiment 3). Error bars represent confidenceinterval computed on data corrected for between-subject variability(Morey,2008), following Baguley (2012, formula 8)recommendations.
At the same time, the absence of increase in interposition errors in groupedsequences is challenging for STM models assuming a hierarchical representationof serial order (Brown etal., 2000; Burgess & Hitch, 1999; Hartley et al., 2016; Henson, 1998; Lewandowsky & Farrell,2008). The ability of these models to account for grouping effects(Frankish, 1985,1989; Hartley et al., 2016;Henson, 1996;Hitch et al.,1996; Maybery etal., 2002; Ng& Maybery, 2002, 2005; Ryan, 1969a, 1969b) relies on the hierarchicalrepresentation of positional information. However, the consequence of usinghierarchical representation of serial order is that any model implementing thatmechanism should predict an increase in interposition errors in groupedsequences, even with shorter sequences. As it is not clear from previousresearch whether the absence of increased interpositions is typical of therecall of 6-item sequences grouped with a 23 structure (see Farrell, 2008; Hitch et al., 1996;Maybery et al.,2002; Parmentier& Maybery, 2008), it is a possibility that this specific groupingstructure represents a particular case. In some positional models (e.g., Brown et al., 2000;Henson, 1998),terminal positions are represented with greater distinctiveness. Thus, thepositional codes of the two groups in a 2-group structure are more distinctivecompared with, for instance, the positional codes between the second and thirdgroups in a 3-group structure. It is then a possibility that a 23 groupingstructure represents a special case in which there is no group at terminalpositions, which then prevents the occurrence of interposition errors due to theincreased distinctiveness between the groups. Further modelling work would berequired to explore this account.
A second aim of Experiment 1 was to discover whether one-trial overshadowing of a sucrose aversion might be found using a novel context. To date, only tastes have been found to produce one-trial overshadowing within a taste-illness paradigm. In a previous study overshadowing of a sucrose aversion was obtained when rats were placed in a novel context (steel cage) 10 min prior to being given sucrose in our standard drinking chambers; however, this effect appeared only after two conditioning trials (Kwok et al., 2012; Experiment 2). A subsequent unpublished experiment produced a similar outcome when rats were placed in the novel context 10 min after drinking the sucrose solution but again overshadowing was revealed only after two conditioning trials. However, if the closer an event is to LiCl injection, the stronger is the one-trial overshadowing effect it produces, this would suggest the possibility that one-trial overshadowing by a context might be detected if the context were experienced late in the sucrose-LiCl interval. Therefore, as shown in Table 1, the fourth group in this experiment was a Context group, whereby these rats were placed in a novel context for 5 min immediately before injection. This timing was chosen with the aim of maximizing the possibility of obtaining overshadowing by a context.
The second type of explanation of overshadowing is one assuming that, even with multiple trials, the acquisition of an association between a target stimulus and some outcome is unaffected by the presence of other stimuli. The latter are important only when the target stimulus is tested. At this point the response to the target stimulus is reduced to the extent that associated stimuli predict the same outcome (e.g., Denniston et al., 2001). This explanation of overshadowing provided by comparator theory is based on overshadowing procedures involving simultaneous compounds, so that the comparison that takes place at test depends on the formation during conditioning of within-compound associations between the target and overshadowing stimuli. Then, at test the target evokes the outcome both directly and indirectly via associative links with the overshadowing stimulus; the stronger the latter evokes the outcome, the weaker the conditioned response to the target stimulus.