Embedding primes in a person’s speech and gestures can influence people’s decision-making


A pair of psychologists at Goldsmiths, University of London has found that embedding primes in a person’s speech and gestures can influence people’s decision-making.

In their paper published in Proceedings of the National Academy of Sciences, Alice Pailhès and Gustav Kuhn describe experiments they conducted with volunteers and primes and what they learned from them.

In both psychology and magic circles, primes are known as actions or words that unconsciously influence the thinking of another person One example is a policeman interrogating a witness tapping his ring while inquiring about jewelry a suspect might have been wearing.

It is a technique magicians have used for years. They prime a person or audience by giving them subtle verbal or physical clues to get them to choose a number during a guessing trick, or a card during a card trick.

In this new effort, the researchers tested the practice to see if it actually works.

The experiments involved asking volunteers to watch a live or taped performance of a person who, unbeknownst to them, was trying to prime them.

The research began with 90 volunteers who were split into two groups. One group watched Pailhès (who is also an amateur magician) perform a live magic act – the other group watched a video version of the same act on a laptop.

The act consisted of attempting to get the crowd of observers to pick a predesignated card – the three of diamonds.

As part of the routine, she mimicked an act by British illusionist Derren Brown in which he asks a member of the crowd to mentally transmit the correct card to the others in the crowd. He also asks the crowd to think about a bright, vivid colored card (more descriptive of a red card than a black one).

He also mimes the shape of a diamond with his hands and asks the audience to think of the little numbers at the corners (ruling out double digits and face cards) and even draws the number “3” in the air with his hands – and asks the audience to imagine things in the middle as he says, “boom, boom, boom.”

After mimicking this act, the researchers asked the audience members to write down a card by suit and number, which they turned in.

Inspection of the cards showed that 17.8 percent of the audience members chose the three of diamonds – 38.9 percent chose a three of any suit and 33.3 percent chose a diamond of any number, results that were far better than chance.

Consciousness and cognitive control
Some authors have argued that there might be some (cognitive) processes truly bound to consciousness, although this is strongly debated (for reviews see Umilta, 1988; Dehaene and Naccache, 2001; Jack and Shallice, 2001; Mayr, 2004; Hommel, 2007; Kunde et al., 2012).

One of the major candidates for this is cognitive control, a general term for cognitive functions that allow us to rapidly and flexibly adapt our behavior when necessary. Cognitive control functions include error detection and correction mechanisms, conflict resolution, response inhibition, and task-switching.

These functions are all strongly associated with the prefrontal cortex, which many consider pivotal for generating awareness (for reviews see Rees, 2007; Dehaene and Changeux, 2011; Lau and Rosenthal, 2011).

Interestingly, some cognitive control processes can be activated by unconscious stimuli. To our knowledge, the first to show that some control processes can be initiated fully automatically and unconsciously were Eimer and Schlaghecken (Eimer and Schlaghecken, 1998; Eimer, 1999).

In an impressive set of studies, they showed that unconscious (masked) arrow primes initially facilitated responses, but can also inhibit responses in certain circumstances.

In their tasks, subjects generally have to respond to a target-arrow (e.g., ») that can be preceded by a congruent (») or incongruent («) masked prime-arrow. When the interval between the prime and target is short (e.g., 50 ms), subjects respond faster and make fewer errors to congruent than to incongruent trials, as might be expected.

However, crucially, when the delay between prime and target was increased (>100 ms), there was no response facilitation but rather automatic inhibition of these responses. This led to the counterintuitive observation that response times (RTs) were faster and error rates lower to incongruent trials compared to congruent trials (note that part of the effect might be explained by lower-level stimulus characteristics, see Lleras and Enns, 2004; Jaskowski and Przekoracka-Krawczyk, 2005; Schlaghecken and Eimer, 2004).

More recently, automatic inhibition paradigms have been combined with brain-imaging tools and the results suggest that automatic inhibition relies on activity in the caudate and thalamus (Aron et al., 2003) as well as the supplementary motor areas (Sumner et al., 2007).

Recent studies have demonstrated the possibility to initiate more “voluntary” forms of response inhibition unconsciously, as studied by using the Go/No-Go task and the stop-signal paradigm (Hughes et al., 2009; van Gaal and Lamme, (in press)).

In these tasks, subjects are required to inhibit an already initiated (stop task) or planned response (Go/No-Go task). To illustrate, in one of these experiments, subjects were instructed to respond as fast as possible to the direction of an arrow (the go stimulus), but to withhold this response when the word “STOP” (the “stop stimulus”) was presented briefly and quickly after the go-arrow (Figure ​(Figure1A).1A).

However, when another word (e.g., “BLUF,” the “go-on stimulus”) was presented, subjects had to continue responding to the direction of the go-arrow. Crucially, the visibility of the stop/go-on stimulus was manipulated by presenting it in between random letter masks.

Therefore, on some trials these stimuli were clearly visible, whereas on other trials they were not. Behaviorally, subjects slowed down their responses to unconscious stop-signals (compared to unconscious go-on signals), as if the STOP signal was briefly processed but not enough to cause a full response inhibition.

When electrophysiological responses to unconscious stop- and go-on signals were compared, a cascade of neural events could be observed, starting early at occipital electrodes, swiftly progressing to fronto-central (the N2 ERP component) and centro-parietal electrode sites (the P3 ERP component), later in time (Figure ​(Figure1B).1B).

Interestingly, in the conscious condition the magnitude of the N2 ERP component was correlated with the efficiency of inhibitory control across subjects (the stop-signal reaction time) and with the magnitude of slowdown to unconscious stop-signals.

Thus, the N2 ERP component likely reflects the initiation of inhibitory control, irrespective of the conscious awareness of the control-initiating stop-signal. The frontal origin of this effect has been confirmed by source reconstruction of the EEG signals (van Gaal et al., 2008) as well as by fMRI (van Gaal et al., 2010b), in similar tasks.

In fMRI, RT slowing to unconscious No-Go signals was associated with focal activations in the pre-SMA and inferior frontal cortices, bordering anterior insula (van Gaal et al., 2010b), whereas response inhibition to visible No-Go signals was related to large scale activation in a typically observed fronto-parietal “inhibition network” (Aron, 2007; Simmonds et al., 2008).

The strength of activation in the unconscious inhibition network was correlated with the extent of slowdown to unconscious No-Go signals across subjects, suggesting that this activation is functional in a sense that it is related to behavioral effects of cognitive control.

An external file that holds a picture, illustration, etc.
Object name is fnhum-06-00121-g0001.jpg
Figure 1
(A) Task-set-up. (B) Electrophysiological processing across time of a masked stop-signal (the word “STOP”) compared to a control “go-on” condition (e.g., the word “BLUF”). Three neural events can be distinguished: (1) an early event at occipital electrodes, (2) a middle event at fronto-central electrodes (The N2 ERP component), and (3) a late event at centro-parietal electrodes (The P3 ERP component). Adapted with permission from van Gaal et al. (2011).

Recent results suggest that several cognitive control functions other than response inhibition can be triggered by unconscious or unnoticed stimuli (for a recent review see van Gaal and Lamme, (in press)), including task-set preparation (Mattler, 2003; Lau and Passingham, 2007; van Opstal et al., 2010; de Pisapia et al., 2011; Reuss et al., 2011; Zhou and Davis, 2012), conflict detection/resolution (Ursu et al., 2009; D’Ostilio and Garraux, 2012) (but see Dehaene et al., 2003; Bruchmann et al., 2011), motivation (Pessiglione et al., 2007; Aarts et al., 2008; Custers and Aarts, 2010) and error detection (Nieuwenhuis et al., 2001; Hester et al., 2005; Klein et al., 2007; O’Connell et al., 2007; Belopolsky et al., 2008; Cohen et al., 2009; Pavone et al., 2009; Dhar et al., 2011) (but see Woodman, 2010).

To illustrate, Lau and Passingham (2007) cued participants consciously to perform either a phonological or semantic judgment on an upcoming word. This conscious instruction cue was always preceded by a conscious or unconscious prime associated with the same or the alternative task (congruent vs. incongruent trials) (see also Mattler, 2003 for a behavioral version of this experiment). When participants were unconsciously primed to perform the phonological task, there was increased activity in a cortical network associated with this task (premotor cortex) and decreased activity in the cortical network associated with the semantic task (inferior frontal cortex and middle temporal gyrus), and vice versa.

These results demonstrate that task-related neural networks, incorporating prefrontal cortex, can be modulated unconsciously. Further, the authors showed that unconscious primes triggered stronger activity in the dorsolateral prefrontal cortex compared to conscious primes, irrespective of the specific task being cued. Recently, Zhou and Davis (2012) went one step further and demonstrated that this effect was not caused by low-level perceptual priming and could still be observed when the unconscious cue was not part of the consciously instructed task-set.

Although it has repeatedly been observed that the strength of unconscious information processing increases considerably with practice and learning (Damian, 2001; van den Bussche et al., 2009; van Gaal et al., 2009), it has been shown that strong stimulus-response bindings are not a prerequisite for subliminal processing to occur (for a meta-analysis see van den Bussche et al., 2009).

However, primes that are also included as targets (“repeated primes”) have a stronger impact and might affect motor responses earlier (have a faster time-course) than primes that are not included as targets (“novel primes”) (Finkbeiner and Friedman, 2011).

In fact, also for higher-level cognitive control processes, such as response inhibition, stimulus-response mappings can be flexibly changed without abolishing unconscious priming effects.

In a task in which a masked stimulus (diamond or square) could be associated with either a Go or No-Go response, but the specific mapping of stimuli onto these actions varied on a trial-by-trial basis (by virtue of a pre-cue), it was recently demonstrated that the same unconscious stimulus could have a substantially different effect on behavior and (prefrontal) brain activity depending on the rapidly changing task-context in which it was presented (Wokke et al., 2011).

In conclusion, several “high-level” (prefrontal) cognitive functions, such as response inhibition and task-switching, have been observed to be influenced and modulated by subliminal stimuli.

These activations seem truly functional, because they are associated with behavioral indices of cognitive control. In the next section, we will discuss the influence of top-down factors (e.g., attention, task-set) on the extent of subliminal information processing and whether subliminal information can initiate top-down cognitive task-sets itself.

Conscious awareness and top-down cognitive control
Traditionally, it has been assumed that unconscious processes were rather automatic, inflexible, and independent of top-down cognitive control (see Hommel, 2007; Kiefer et al., 2012 for reviews).

However, accumulating evidence shows that unconscious information processing is not fully automatic, but can be modulated by several top-down cognitive and attentional factors.

Overall, the instructed task-set and subjects’ strategy strongly affects the strength, direction and depth of subliminal information processing (Kunde et al., 2003; Greenwald et al., 2003; Ansorge and Neumann, 2005; Kiefer and Martens, 2010; Al-Janabi and Finkbeiner, 2011; O’Connor et al., 2011).

For example, the top-down instructed task-set, e.g., either to read aloud a visible target word or to categorize it as representing natural or artificial objects, can change the processing route taken by an unconscious (masked) word preceding the target word (Nakamura et al., 2007).

Along similar lines, Kiefer and Martens (2010) recently showed that the N400 ERP component to unrelated prime-target pairs (e.g., masked word “chair” followed by a visible target word “leaf”), compared to related prime-target pairs (e.g., masked word “chair” followed by visible target word “table”), was enhanced when a semantic task-set was induced by a visible cue presented immediately before each trial and was attenuated by a perceptual task-set (see also Martens et al., 2011).

Further, attended subliminal stimuli have a stronger impact on behavior than unattended subliminal stimuli, and this is the case for spatial attention (Kentridge et al., 1999, 2004, 2008; Sumner et al., 2006; Bahrami et al., 2008a; Marzouki et al., 2008; Finkbeiner and Palermo, 2009), temporal attention (Naccache et al., 2002; Kiefer and Brendel, 2006; Fabre et al., 2007) and during attentional load (Bahrami et al., 2008b; Martens and Kiefer, 2009).

Task-relevant (attended) stimuli are processed stronger than task-irrelevant (unattended) stimuli, even when unconscious. Ansorge and Neumann (2005) showed that task-relevant prime features (e.g., shape) affected responses to the target only when the shape dimension was response relevant, but not when this feature was task-irrelevant, for example when the color of the target determined the required response (see also Tapia et al., 2010).

We recently explored the role of task-relevance of subliminal information using EEG in a task in which subjects had to respond as fast as possible to a black Go annulus, unless it was preceded by a briefly presented gray circle (the no-go stimulus).

Due to variations in the SOA between the No-Go circle and Go annulus, on some trials the No-Go circle was perceived consciously, whereas on others it was not. On the current trial, unconscious No-Go circles activated prefrontal control networks (van Gaal et al., 2008), and the extent to which correlated strongly with the amount of RT slowing to these stimuli. Crucially, exactly the same subliminal gray circle did not activate the PFC when it was task-irrelevant, but presented in a highly similar task-context (although it yielded similar early visual responses).

This result highlights that the processing route taken by an unconscious stimulus strongly depends on task-relevance (and attention to the stimulus), and that task-irrelevant subliminal stimuli probably decay rapidly while progressing up in the cortical hierarchy.

Recently, it has been observed that, under some conditions, cognitive control processes can still be influenced by subliminal stimuli presented outside the direct focus of spatial attention (Rahnev et al., 2012).

The role of attention and other top-down factors for unconscious information processing might depend on type of information to be processed. Recent research suggests that attention might be more crucial for “neutral” stimuli (e.g., numbers: Naccache et al., 2002) than for emotional, arousing or “evolutionary relevant” stimuli.

To illustrate, Finkbeiner and Palermo (2009) have found that masked pictures of face stimuli produced priming regardless of whether they were spatially attended (however, this was not the case for subliminal eye-gaze cues: Al-Janabi and Finkbeiner, 2011).

In contrast, other non-face stimuli (animals, vegetables) only produced subliminal priming when attended (see also Harry et al., 2012). However, although it seems that the threshold for conscious access is lower for emotional stimuli (Gaillard et al., 2006) and that these produce stronger priming (Brooks et al., 2012), also emotional information processing does not seem to be fully automatic and is also modulated by top-down “attentional sensitization,” at least to some extent (Kiefer et al., 2012).

In fact, even when emotional pictures (e.g., faces) are presented fully consciously their depth and extent of processing seem to be facilitated by attentional factors (Pessoa et al., 2002, 2003).

Attention itself can also be attracted unconsciously (for review see Mulckhuyse and Theeuwes, 2010), for example by threatening (Lin et al., 2009), emotional (Vuilleumier and Schwartz, 2001; Brooks et al., 2012), erotic (Jiang et al., 2006), or socially relevant stimuli (Sato et al., 2007), but also by lower-level stimulus attributes, such as gamma flicker (Bauer et al., 2009) and stimulus orientation (Rajimerhr, 2004).

Recently, it has been shown that individual differences in attentional bias to masked fearful faces are related to gray matter volume in the anterior cingulate cortex (Carlson et al., 2012), suggesting that these attentional effects are truly top-down mediated.

The literature reviewed above illustrates that consciously instructed task-sets and strategies as well as attentional factors strongly influence the processing of subliminal stimuli in various ways.

At present, it is still an open and important question whether top-down task-sets can also be triggered by subliminal information. Several studies have reported so-called “top-down context effects.”

In these experiments, subjects generally perform a masked priming task consisting of congruent and incongruent prime-target pairs. The crucial manipulation in such experiments is the ratio of congruent and incongruent trials within experimental blocks. In blocks in which the prime direction does not predict the direction of the upcoming target (50% congruent and 50% incongruent trials) subjects are generally faster to congruent than to incongruent trials.

However, several experiments have consistently revealed that the impact of conflicting stimuli on behavior is larger when incongruent prime-target pairs are infrequent (∼20%) compared to when these are frequent (∼80%), at least when conflicting stimuli are presented consciously (for review see Desender and van den Bussche, 2012).

In fact, the effect might even completely reverse in such a way that responses to incongruent prime-target pairs are faster than to congruent pairs (Merikle and Joordens, 1997; Daza et al., 2002), because subjects are able to strategically use the prime information to predict the upcoming target category.

Even for conscious trials this might take some time (∼400 ms), suggesting that these strategic effects take some time to build up (Ortells et al., 2003). These conscious strategic effects were recently only observed for spatially attended stimuli, but not for unattended ones (Ortells et al., 2011). At present it is still disputed whether such context effects depend on the conscious awareness of the primes, because several studies have reported an absence of congruency effects when the conflicting stimuli were presented subliminally (Merikle and Joordens, 1997; Daza et al., 2002; van den Bussche et al., 2008; Heinemann et al., 2009).

However, other studies have shown that context effects also apply to unconscious prime stimuli (Jaskowski et al., 2003; Bodner and Masson, 2004; Wolbers et al., 2006; Klapp, 2007; Bodner and Mulji, 2010). Interestingly, these context effects initiated by subliminal primes might be related to increased connectivity between the pre-SMA and stimulus-related (LOC) and motor-related (putamen) brain areas (Wolbers et al., 2006), suggesting that the pre-SMA plays a role in the strategic control over the processing of subliminally presented conflicting stimuli.

Several authors have noted that it is important to examine whether these context effects are truly unconscious, at all processing levels, and which part of the effect might be explained by meta-cognitive (conscious) processes.

For example, subjects might become aware of the increased error rate, experienced “difficulty” or “effort” on blocks with high numbers of conflicting trials and thereby might strategically adapt their response strategy or attentional focus (Jaskowski et al., 2003; Kinoshita et al., 2008, 2011, for a more extensive discussion of this issue see Desender and van den Bussche, 2012 and below). Therefore, it is still an open question whether top-down context effects can also be initiated by unconscious stimuli (Dehaene and Naccache, 2001).

Heinemann et al. (2009) studied the role of conflict awareness in a slightly different way, namely by examining the role of context on conflict frequency effects, also referred to as the context-specific proportion congruent effect (see also Crump et al., 2006).

They performed a typical masked priming task in which subjects had to categorize target numbers as being larger or smaller than 5. A target was always preceded by a masked prime number that could be congruent or incongruent to the target.

Crucially, just before the presentation of the prime-target pair they presented a colored rectangle at the background that determined the congruency context (the colored rectangle disappeared upon presentation of the response feedback).

One color was consistently associated with a low interference context (80% congruent trials, 20% incongruent trials), whereas another color was associated with a high interference context (20% congruent trials, 80% incongruent trials).

As predicted, for weakly masked primes (visible) the congruency effect (RT incongruent—RT congruent) was significantly smaller in the high interference context than in the low interference context (32 vs. 54 ms).

Crucially, they showed that these context-specific congruency effects were absent for strongly masked (poorly visible) trials. The authors concluded that context-specific congruency adaptation requires conscious representation of the conflicting information.

Interestingly, previous work suggests that, even when using visible stimuli only, subjects do not have any explicit awareness of the congruency manipulation in similar tasks (Crump and Milliken, 2009).

Therefore, it has been suggested that context-specific congruency effects might not depend on explicit knowledge of the congruency proportions, but might require sufficiently strong (i.e., conscious) representations of the prime, target and context (Kunde et al., 2012).

In a recent study, van Opstal et al. (2011a) took a somewhat different approach and demonstrated that context effects might indeed be initiated by subliminal primes. In their task, subjects had to indicate whether two target numbers (e.g., 3–3) were the same or different.

These target numbers were always preceded by a masked (subliminal) prime. The crucial comparative prime consisted of a capital letter and a lower-case letter (A-a). In one experiment these primes were mixed with primes consisting of two completely different letters (a-D, the low-similarity context, Figure ​Figure2A),2A), whereas in another experiment they were mixed with primes consisting of exactly the same letters (a-a, the high-similarity context, Figure ​Figure2B).2B).

In the low-similarity context where a-A primes were relatively similar to a-a primes (compared to a-D primes), a-A primes facilitated a “same” response to the targets (Figure ​(Figure2C).2C). On the other hand, the same prime (a-A) presented in the high-similarity context (containing a-a primes) was relatively different and indeed facilitated a “different” response to the targets (Figure ​(Figure2D).2D).

Importantly, RTs were equated across conditions and, therefore, could not (directly) explain the observed effects. This may be an important step in further pushing the boundaries of unconscious information processing and opens the possibility that also a top-down task-set may be enabled unconsciously (Dehaene and Naccache, 2001).

An external file that holds a picture, illustration, etc.
Object name is fnhum-06-00121-g0002.jpg
Figure 2
(A,B) Task-set-up. (C,D) Response times for the low-similarity and high-similarity context for the different prime types and prime-target congruency. Adapted with permission from van Opstal et al. (2011a).

By now it is well established that subliminal information processing (e.g., its depth, extent, and direction) is influenced by several top-down cognitive functions, such as attention, task-set and strategy.

However, whether top-down context effects themselves can be initiated or affected by subliminal stimuli is still under scrutiny (see also Kunde et al., 2012). In this respect, the underlying mechanisms, boundary conditions and role of awareness in blockwise congruency effects (Desender and van den Bussche, 2012) and context-specific proportion effects (Heinemann et al., 2009) are interesting avenues for future experimentation. Next, we will discuss another crucial and disputed aspect of subliminal information processing: its alleged short-lived nature.

Consciousness and decision-making
The literature we reviewed so far shows that unconscious information can affect high-level processes, and might even act on aspects of cognitive control and (working) memory.

Lastly, we will discuss studies that investigated whether and how unconscious information can be accumulated across time or space for perception and decision-making. Active information integration is considered one of the hallmarks of consciousness by many contemporary models of consciousness (Tononi and Edelman, 1998; Engel and Singer, 2001; Crick and Koch, 2003; Baars, 2005; Seth et al., 2008).

Several recent studies have observed that, under some conditions, subliminal information can be accumulated and integrated spatially across the visual field (van Opstal et al., 2011b).

In one of van Opstal’s experiments, subjects were presented with 4 spatially separated numbers (primes) that were preceded and followed by masks that prevented conscious perception of the primes.

A target, also consisting of four digits, followed the prime rapidly and subjects had to indicate whether the mean of the 4 target digits was more or less than 5.

Interestingly, the mean of the subliminal primes affected RTs and accuracy to target responses, suggesting that so-called “ensemble statistics” might be extracted unconsciously.

Other recent evidence also suggests that multiple unconscious stimuli can be integrated across space, for example when visual scenes are presented in the absence of awareness because of continuous flash suppression (Mudrik et al., 2011).

Also, expert chess players (but not novices) are able to extract whether a subliminal (masked) simplified chess configuration entails a checking configuration or not.

However, this was only the case for highly familiar chess configurations, and was not present in a task that required the integration of local features, namely field color (black or white) and chess piece (rook or knight).

This suggests that experts have created chunks of common chess configurations in long-term memory (which novices have not) and, therefore, that they might not actively have to integrate individual stimulus features (Kiesel et al., 2009).

Generally, the extent of practice might be crucial and partly explain why evidence in the field is somewhat mixed. Others have shown that the integration of local features into global shapes does require stimulus awareness, for example when stimuli are rendered invisible due to counter-phase flickering of stimulus contrast (Schwarzkopf and Rees, 2010).

Unconscious information also seems to be integrated or accumulated across time, at least to some degree. Previous studies have shown that subliminal information can be accumulated linearly over a few hundreds of milliseconds (Jaskowski et al., 2003; Vorberg et al., 2003; Wentura and Frings, 2005; Del Cul et al., 2007; Frings et al., 2008). For example, Jaskowski et al. (2003) nicely showed that increasing the number of primes presented before a target increases the behavioral priming effect.

In their task, subjects were required to respond to the spatial location of a square with horizontal gaps presented together with a square without such gaps. Targets could be preceded by either 1, 2, 3 or 4 primes (presented for 35 ms each) which were smaller copies of the target.

Because the squares in every next stimulus were slightly larger than the previous ones, they masked the preceding stimulus. They showed that each of the 4 primes had an influence on the response to target, and that with increasing number of primes the priming effect was larger. Similarly, Vorberg et al. (2003) have shown that when the time between prime and target is increased (from 14 to 86 ms in steps of 14 ms) the behavioral priming effect increases monotonically.

Subjects had to respond to the direction of a metacontrast target arrow that was preceded by a smaller version of it. Importantly, because the stimuli were presented outside the focus of attention (below and above fixation), there were no SOA-related changes in prime awareness (see also Schmidt et al., 2010). Together, these results suggest that subliminal information can be accumulated over short periods of time (<150 ms) and increasingly impact behavior.

However, while the accumulation of information may be possible irrespective of the level of awareness over short periods of time, recent studies have shown that awareness might play an important role when the time across which information has to be accumulated is increased. de Lange et al. (2011) performed a task in which subjects had to accumulate multiple pieces of evidence across 1.5 s.

On each trial, subjects were presented a stream of five arrows, each of which could point to the left or right with equal probability. They had to quickly decide on the direction of the majority of arrows, guessing if necessary (Figure ​(Figure4A).4A).

The strength of the evidence could range from one (low evidence, e.g., two left and three right arrows) to five (high evidence, e.g., five right arrows). The visibility of the arrows was manipulated by masking them with an effective metacontrast mask leading to arrows near the threshold of awareness (low visibility condition) or with an equiluminant but less effective “pseudo” mask (leading to high visibility). On each trial, all arrows were either of low- or high visibility.

Qualitative differences in perception were confirmed by objective as well as subjective discrimination measurements (leading to low vs. high visibility arrows, instead of conscious vs. unconscious arrows).

Importantly, stimulus and mask duration were identical for both conditions, which allowed the comparison of behavioral performance of evidence accumulation (and the underlying neural responses) without confounding stimulus visibility with basic task parameters (e.g., signal strength) (Lau, 2008; Francken et al., 2011).

Behaviorally, subjects were able to accumulate evidence over time for both visibility conditions (Figure ​(Figure4B).4B). However, there were marked qualitative differences in how information was accumulated for the different levels of awareness.

First, decision-making speed was modulated by the amount of accumulated evidence, but only for high-visible stimuli (Figure ​(Figure4C).4C). Second, once enough evidence had been gathered, participants strategically reduced the impact of new incoming stimuli (Figure ​(Figure4D).4D).

Crucially, by using the same stimulus parameters but now in a masked priming task, it was observed that the amount of bottom-up information provided by the arrows was the same for both conditions, as reflected in an equal size of the behavioral priming effect for both visibility conditions (Figure ​(Figure4E).4E).

Thus, although unconscious evidence may be accumulated in a linear fashion, i.e., adding and subtracting new information without any regard to the history of prior accumulated evidence, non-linearities in evidence accumulation (for example, reducing the weight of new information under conditions of high certainty, Kiani et al., 2008; de Lange et al., 2010) may be present only for fully consciously perceived information.

This qualitative difference (linear vs. non-linear integration) was also observed in concurrently measured neural recordings: occipito-parietal regions that were involved in the accumulation of the sensory evidence showed a “linear” stereotypic response when presented with near-threshold information, but modulated their activity strategically during the task for clearly visible information.

These results suggest that the level of awareness of information changes decision-making: while accumulation of evidence is already possible for low visibility information, high visibility allows evidence to be accumulated up to a much higher-level, leading to important changes in strategic top-down decision-making.

More information: Alice Pailhès et al. Influencing choices with conversational primes: How a magic trick unconsciously influences card choices, Proceedings of the National Academy of Sciences (2020). DOI: 10.1073/pnas.2000682117


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.