On the limits of evidence accumulation of the preconscious percept
Introduction
It is uncontroversial to believe that memories for a stimulus are strengthened through repeated experience of that stimulus. Indeed, Endress and Potter (2014) have explicitly shown this to be the case, even for briefly fixated objects. However, a key question is the role of conscious perception in this process. In this respect, the scenario presented in Case 1 in Fig. 1 is not controversial; that is, if a stimulus is consciously perceived, some sort of “trace” (whether activation-based or synaptic) of that stimulus would typically form after the moment of awareness. Furthermore, such a trace could accumulate over repeated presentations, i.e. there would be evidence accumulation.
What happens though, in the case of stimuli that do not reach awareness is less clear. That is, would it be possible for the trace of an item to strengthen, i.e. for evidence to accumulate, through repeated presentation if none of those presentations induced a conscious percept, such that conscious perception becomes more likely with further presentations, as in Case 2 of Fig. 1. In contrast, it could be the case, as shown in Case 3, that stimuli that do not reach a threshold of awareness do not leave a trace that accumulates over multiple exposures.
This is the question we consider in this paper, viz, we seek to determine whether the brain behaves as Case 2 or Case 3 of Fig. 1. In other words, do representation traces for stimuli registered below the awareness threshold, dissipate back to baseline so quickly that there is effectively no evidence accumulation for the preconscious percept? (How these different modes of evidence accumulation relate to relevant phenomena in the literature, such as subliminal priming, is considered in the Discussion section.)
To answer this question, one needs a means to present a lot of stimuli in such a way that many do not cross the awareness threshold. The natural way to do this is with Rapid Serial Visual Presentation (RSVP), for which it is known that only a small subset of the presented stimuli are reportable, or indeed recognisable (Bowman et al., 2013; Bowman, Filetti, Alsufyani, Janssen, & Su, 2014; Potter, 1976).
In fact, there are previous studies that have considered the progressive strengthening of memories with RSVP (Albrecht & Vorberg, 2010; Endress & Potter, 2014; Subramaniam, Biederman, & Madigan, 2000). For example, Endress and Potter (2014) reported that images (and words) presented more often across a number of Rapid Serial Visual Presentation (RSVP) streams were recognized more accurately in a final recognition test. Given that, as previously discussed, in RSVP studies, participants often show very poor recognition performance, Endress and Potter’s findings are open to the interpretation that some items leave memory traces that gain strength through repetition, despite not being consciously perceived. That is, although we will ultimately argue against this position, Endress & Potter’s findings open the possibility that Case 2 of Fig. 1 obtains and memories accumulate through repetition for the pre-conscious percept.
However, some studies have reported that most stimuli presented in RSVP do not display consolidation/evidence accumulation through repeated exposure. Subramaniam et al. (2000) presented participants with RSVP streams of drawings of objects. Participants were instructed to search for a target image. Crucially, some of the non-target pictures were presented 15 times on average before becoming a target. The results revealed an absence of a repetition effect; that is, participants were not better at detecting the targets that had previously been repeated relative to those presented once. In a similar vein, Bowman et al. (2014) presented participants RSVP streams of first names and they were instructed to search for a Fake Name (a name they were pretending was their name). In addition, in Experiment 3, they were instructed to search for frequently presented names; that is, they had to search for names simply on the basis that they were repeated. Importantly, these repeating names were presented as often as the Fakes (up to 50 times). The behavioural results (recall and recognition test at the end of the experiment) indicated that participants found it very hard to identify the repeated names; and consistent with this, in ERP findings, the Fake generated a clear P3 that was absent for the repeating name. As in Subramaniam et al. (2000), participants were actively searching for an additional item (Target in Subramanian et al, Fake in Bowman et al), which arguably could hinder the encoding process of repeating items.
Taken together, the above described findings are far from offering a coherent picture. On the one hand, some studies raise the possibility that evidence accumulation is possible for non-retrievable (and thus not consciously perceived) stimuli presented in RSVP.1 This would suggest that items naturally elicit (graded-strength) memory traces. On the other hand, other studies suggest that the capacity to form memory traces from the fleeting representations of items in RSVP is very limited. That is, despite the serial presentation of items in RSVP, unless an item is processed to the point that it reaches a state of awareness, evidence for it would not accumulate.
One can view the present study as shedding light on the nature of the discrepancies between the described findings. If the presentation of stimuli in RSVP results in memory representations of gradual strength, these representations may increase in strength with every repetition, facilitating retrieval. On the other hand, the encoding of items could follow a bottleneck behavior: a small number of stimuli would (enter consciousness and) be stored in stable WM representations, while the vast majority would not reach that stage. Importantly, the “missed” stimuli would not benefit from successive repetitions, since residual information would not survive the presentation of new items.
In the previous studies that have considered the progressive strengthening of memories with RSVP (Subramaniam et al., 2000; Endress & Potter, 2014), a repeating stimulus occurred incidentally in RSVP streams through the course of an experiment. Then a recognition test on the stimulus was inserted after a certain number of repetitions. This previous work, though, was not specifically focussed on preconscious evidence accumulation, which is our interest. That is, they did not probe memory in such a way that they could identify the first time a stimulus was seen as repeating, leaving the possibility that their recognition reports could have arisen after at least some previous presentations were consciously perceived, i.e. Case 1 in our Fig. 1. As a result, we have had to employ a somewhat different experimental paradigm to these previous studies. In particular, we could not rely on recognition memory tests at the end of each of a number of RSVP streams, since such a test could reveal the identity of a repeating (target) stimulus whether it had or had not been perceived in a stream to that point, thereby consciously priming its future perception. This would confound any test of an intrinsic below threshold build-up of evidence with repeated presentation.
We are interested in isolating the first instance at which a (non-primed) repeating item is consciously perceived as repeating. To obtain such a test, we have run an RSVP repetition experiment, where we instruct participants that a repeating item will be presented, but we do not identify it and participants are required to search for it simply on the basis that it repeats.
With this approach, we can test what turns out to be the key property for us, which is that (first) detection of repetition is invariant to the number of prior presentations, with the following procedure.
- 1)
We determine the first instance at which an item is seen as repeating.
- 2)
We assess whether the probability of this first instance is invariant to the number of prior repetitions; by determining the conditional probability of seeing a repetition, given that it has not been seen as a repetition before.
If this conditional probability, which we call the first seen as rep. probability, is indeed invariant across repetitions, there is no evidence accumulation before a repetition is first seen. This can be illustrated with reference to Fig. 1, where the accumulation inherent to Case 2 would ensure that the probability of first seeing (and then seeing as a repetition) would increase with each sub-threshold registration of a stimulus. This is because the distance to threshold would be reducing on each registration. In contrast, in Case 3, the probability of first seeing (and then seeing as a repetition) would not change with (below threshold) registration of a repeating stimulus. This interpretation of Cases 1 and 2 is confirmed in simulations in the appendix.
More specifically, we have designed two different tasks using RSVP of streams of words: (1) a Repetition task and (2) a Detection task. In the first of these, the repetition task, participants were instructed to search for a word that was repeated across trials/ streams. They were informed that they were going to see several streams of words and they must search for the repeated word. At the end of each trial/ stream, they had to answer the question “Which one was the repeated word?”. We varied the presentation time in two experiments (from a SOA of 17 ms–533 ms). In this task, we tested to what extent participants can detect repeated items inserted in different RSVP trials. This enabled us to examine to what extent contents generated in RSVP can accumulate evidence through repetitions, with a key test being invariance to prior presentation of the first seen as rep. probability. To do this, we examined whether the number of presentations of a word increased the probability of first detecting the repeated word.
While, as we have discussed, the experimental paradigm we employ here is somewhat different to that used in (Subramaniam et al., 2000; Endress & Potter, 2014), we believe our experiment and theirs are comparable. In particular, our experiment can be seen as a generous test of the evidence accumulation question, since we are instructing participants to look for repetitions. That is, if evidence does not accumulate preconsciously for repeating items, when participants are explicitly instructed to look for such items, it seems unlikely that is would incidentally, when participants are not instructed to look for them. Incidental build up is the approach in the Subramanian, et al and Endress & Potter experiments.
The results of the Repetition Task were compared with those of a Detection task, which effectively served as a baseline to compare against. In the Detection task, participants were instructed to search for pre-specified target words. This enabled us to explore to what extent participants can search for stimuli in RSVP on the basis of their perceptual properties. With the comparison of the two tasks, we aimed to illustrate the time-course of two fundamentally different types of search: one based on the perceptual features of task relevant items (Detection) and the other based purely on frequent occurrence (Repetition). Our findings will confirm that, consistent with previous work (Bowman et al., 2013; Potter, 1976), the brain is exceptionally good at searching for pre-specified items, as per our detection task. In contrast, searching on the basis that an a priori unknown item occurs frequently, as per our repetition task, is much harder.
Section snippets
Participants
21 undergraduate students of the University of Birmingham took part in Experiment 1 in exchange for course credits. All were right handed, native English speakers and had normal or corrected-to-normal vision. The experiment conformed to British Psychological Society criteria for the ethical conduct of research and ethical procedures of the School of Psychology at the University of Birmingham.
Materials
36 English nouns were selected from the English Lexicon Project database (Balota et al., 2007) to serve
Repetition and detection task
Although the main piece of evidence we are seeking to identify is the estimation of the (accumulation) repetition effect in RSVP (which is described in the next section), the results described here – comparing the d’ score of the Detection and Repetition task – give a global picture of the time-course of participants’ performance in both task.
Discussion
In two studies, participants were presented streams of words at different presentation rates, while performing a detection or a repetition task. The repetition task required participants to search for repeated words across several streams (different trials). The detection task instructed participants to search for pre-specified target words within each trial. Participants were able to complete both tasks, with performance being considerably better in the easier detection task. More importantly,
Funding
This work was supported by the National Science Foundation grant 1734220.
References (22)
The conscious access hypothesis: Origins and recent evidence
Trends in Cognitive Sciences
(2002)- et al.
Consciousness cannot be separated from function
Trends in Cognitive Sciences
(2011) The masked repetition priming effect dissipates when increasing the inter-stimulus interval: Evidence from word naming
Acta Psychologica
(1996)Neural events and perceptual awareness
Cognition
(2001)- et al.
Long-lasting effects of briefly flashed words and pseudowords in ultrarapid serial visual presentation
Journal of Experimental Psychology Learning, Memory, and Cognition
(2010) - et al.
The english lexicon project
Behavior Research Methods
(2007) - et al.
Fitting linear mixed-effects models using lme4
Journal of Statistical Software
(2015) - et al.
Maintenance of non-consciously presented information engages the prefrontal cortex
Frontiers in Human Neuroscience
(2014) Consciousness, accessibility, and the mesh between psychology and neuroscience
The Behavioral and Brain Sciences
(2007)- et al.
Countering countermeasures: Detecting identity lies by detecting conscious breakthrough
PloS One
(2014)
Subliminal salience search illustrated: EEG identity and deception detection on the fringe of awareness
PloS One
Cited by (4)
Is predictive coding falsifiable?
2023, Neuroscience and Biobehavioral ReviewsBreakthrough percepts of famous names
2021, CortexCitation Excerpt :This is consistent with our other experiments on repeating non-salient stimuli across RSVP streams. These found that the probability of a participant “seeing” a repeating non-salient item does not increase with repeated presentations (Avilés et al., 2020) and found no evidence that participants can differentially “see” non-salient repeating items even when their task is to “look for them” (Bowman et al., 2014). Therefore, it seems unlikely that innocent participants would see a repeating non-salient item and result in false detection when they are not even looking for it.
No Subliminal Memory for Spaced Repeated Images in Rapid-Serial-Visual-Presentation Streams
2022, Psychological ScienceModelling the simultaneous encoding/serial experience theory of the perceptual moment: a blink of meta-experience
2022, Neuroscience of Consciousness