Early and Late Stages of Perception Do Not Seem to Be Affected by Motivation Manon Vollebergh
Student number: 10738347 Supervisor: Timo Stein
Number of words: 7838 (excluding the literature list) Number of words abstract: 236
Abstract
This study focused on the question if motivation affects the early and the late stages of perception, by look at whether reward-related information was prioritized in perception. The study was partly a replication of the study done by Raymond and O’Brien (2009) and therefore a value-learning task and an attentional blink recognition task were included to study the late stages of perception. However, this study also included an attentional blink localization task in order to study the early stages of perception and a rating task. Both attentional blink tasks were performed with and without constraints of available attention, by using a short and a long lag. Fifty-one people participated. The results showed that reward and punishment related faces were learned equally well. Moreover, faces were localized and recognized better after the long lag than after the short lag. Valence and expected value did not affect the accurate localization and recognition of faces. Furthermore, exposure also did not affect accurate localization of faces and reward-related faces were recognized equally well after the short and after the long lag. In addition, the results showed that faces were rated equally likable, regardless of their valence. In conclusion, this study was unable to support the idea that motivation affects the early stages of perception and it was also unable to replicate the findings of Raymond and O'Brien (2009), so motivation also does not seem to affect late stages of perception.
Does Motivation Affect Perception, in Both the Early as The Late Stages of Perception? The world is full of visual information and details, which are interpreted by the people living in this world. This interpretation of visual information, is what is called perception. When there are for example two grey rectangles, the difference in lightness between those two can be noticed due to perception (Firestone & Scholl, 2016). However, an important question is whether people perceive the world as it is or if this perceived world is affected by top-down effects, such as our own motivations, emotions, desires, beliefs and expectations. This question has been hotly debated for decades now. Bruner and Goodman (1947) did a pioneering study and were the first ones declaring that top-down effects influence perception. These results were the beginning of the New Look movement in perceptual psychology, which triggered many studies claiming they found influences of top-down effects on perception as well (Firestone & Scholl, 2016). However, the New Look movement eventually became less popular due to methodological and theoretical scrutiny. After that, the main idea became that there is a clear separation between perceptual and cognitive processing (Firestone & Scholl, 2016). Nonetheless, in the last two decades the main idea returned to the New Look understanding of perception and the robust division between perception and cognitive processing was left behind. Nowadays, the increasingly popular idea is that what we perceive is a combination of bottom-up and top-down factors (Firestone & Scholl, 2016). Nevertheless, there is still no consensus about whether top-down effects are influencing perception. It is important to get an answer to this question, because most models in vision science do not consider top-down factors. This means that a revolution in the understanding of perception is in order, when it turns out that top-down effects affect what and whether people see something.
A definition of top-down effects is necessary to answer the question if there are influences of top-down effects on perception. In the current study the following definition is used:
higher-level cognitive states that repeatedly influence perception (Firestone & Scholl, 2016). The current study will focus on motivation as a top-down effect and will look into the question whether motivation affects both the early as the late stages of perception, by looking at whether reward-related information is prioritized in perception.
Previous studies which explored the influence of top-down effects on perception often used an attentional blink, binocular rivalry or breaking-continuous flash suppression task. An attentional blink occurs when two masked targets (T1 and T2) are presented within 500
milliseconds, participants are often unable to report the second target (T2) (Shapiro, Raymond & Arnell, 1997). The idea is that when a participant is highly motivated to see the second target, for example when it is related to reward, that there will be no attentional blink. Hereby, participants will then be able to report the second target (Raymond & O’Brien, 2009. In addition, binocular rivalry occurs when perceptually dissimilar images are presented to different eyes. Those two percepts then start to compete for perceptual dominance. Every few seconds another percept is dominant. It is possible to determine which visual input the brain is selecting for conscious experience, by measuring the length of dominance durations of both percepts (Anderson, Siegel, Bliss-Moreau, & Barett, 2011). A (breaking-)continuous flash suppression task is the same as a binocular rivalry task, but instead of using two images, a rapidly changing Mondrian-like pattern is shown to one eye and an image is shown to the other eye (Tsuchiya & Koch, 2005).
The current study distinguishes between early and late stages of perception. The early stages of perception are the stages in which detection of stimuli takes place, thus in this stage is decided whether or not something is seen. The late stages of perception are the stages in which recognition and discrimination of stimuli takes place, so here is explored what is seen. In this review of previous studies, studies regarding early stages and studies regarding late stages of perception are therefore distinguished. Some of the previous studies gave evidence supporting the
idea that early stages of perception are influenced by top-down effects, but other studies provided non-supporting evidence. For example, Milders, Sahraie, Logan and Donellon (2006) used an attentional blink task to find out if motivation, operationalized by emotional meaning, affects the ability to detect faces. The results suggest that fear conditioning can modulate detection of formerly neutral stimuli. This means that this can be seen as prove for influences of motivation on early stages of perception. However, this study contains a methodological shortcoming. This study was not performance-based, because participants only had to tell whether they saw something but not where they saw it. Therefore, the found results could be due to participants having a lower criterion to say yes for fear-conditioned stimuli than for non-conditioned stimuli, while actually faces in both conditions were perceived equally well. The current study overcomes this problem, by only using neutral faces instead of emotional faces.
Binocular rivalry and breaking-continuous flash suppression tasks are also useful ways of studying possible top-down effects on detection. Anderson, Siegel, Bliss-Moreau and Feldman Barrett (2011) used a binocular rivalry task in their study and demonstrated that faces previously paired with descriptions of negative social behaviours were prioritized for consciousness longer than faces paired with other gossip or non-social information. This study also showed that associating a person with negative social behaviours makes it more likely that a perceiver will consciously see that structurally neutral face. This study shows that motivation, operationalized as negative, positive or non-social information, decides whether something is seen and therefore provides evidence for influences of motivation on the early stages of perception. Rabovsky, Stein and Abdel Rahman (2016) tried to replicate the findings of Anderson et al. (2011) by using a continuous flash suppression task. Rabovsky, Stein and Abdel Rahman (2016) replicated affective knowledge effects on perception of emotional expressions in neutral faces, but did not find evidence for influences of affective knowledge on suppression durations. Moreover, no
effects of affective knowledge were found for newly learned faces. Moreover, Stein, Grubb, Bertrand, Suh and Verosky (2017) also tried to replicate the findings of Anderson et al. (2011) by using a binocular rivalry task and a breaking-continuous flash suppression task. However, they could not replicate the findings from Anderson et al. (2011). The results of the study showed that faces associated with negative behaviours were not prioritized over faces associated with neutral or positive information. However, faces associated with any type of behaviour were prioritized over novel faces. They found the same results across the binocular rivalry task and the breaking-continuous flash suppression task. The results of the binocular rivalry task are more surprising, because it is contrary to the findings of Anderson et al. (2011). However, the study of Anderson et al. (2011) contains a methodological shortcoming, because the task in this study was not performance-based. Therefore, there might be a problem of demand characteristics. Participants had to press a key for the duration that they consciously experienced seeing a house and a different key for the duration when they saw a face. Because of this, the found results could be due to the expectations of the participants about the study and giving the corresponding answer, instead of giving the answer that corresponds to what they actually saw. The current study overcomes this problem by using performance-based tasks, where the stimuli are shown rapidly after each other so that participants do not have time to think about what answer corresponds with their thoughts about the experiment.
The above-mentioned studies taken together show that there is inconsistency regarding the question whether motivation affects the early stages of perception, whereby it is unclear if
motivation affects whether something is seen. The conflicting results might be due to response bias and demand characteristics. The current study will overcome these problems and is therefore an important addition to the already existing literature.
effects on the late stages of perception. This means that these studies gave evidence for influences of motivation on recognition and discrimination of learned stimuli. For example, Wilbertz, Van Slooten and Sterzer (2014) used a binocular rivalry task to study motivational influences on recognition of stimuli. The results showed that perception is affected by reward and punishment in opposite directions. However, this study has a methodological shortcoming. Participants were given a reward during this task, which is a problem because then the motivation to see certain stimuli is only learned during the task. Therefore, the results of the first trials are not able to provide answers to what is attempted to investigate. The current study overcomes this problem, by learning motivation before the attentional blink tasks are performed.
In addition, attentional blink tasks are also used to study the influences of top-down effects on the late stages of perception. For example, Müsch, Engel and Schneider (2012) used an attentional blink task and showed that emotional faces were less affected by attentional blink compared to non-emotional faces. This means that there is an attentional advantage for faces associated with a high motivation, which are operationalized as emotional faces and easy to distinguish stimuli. Moreover, Raymond and O’Brien (2009) also used an attentional blink task and showed that regardless of available attention, recognition was substantially enhanced for motivationally salient stimuli (i.e. stimuli highly predictive of outcomes), compared with equally familiar stimuli that had weak or no motivational salience. This effect was found regardless of valence (reward or punishment). Moreover, recognition was also enhanced for reward-associated stimuli.
The studies which have just been discussed show that there is consistency regarding the question whether motivation affects the late stages of perception, whereby it is clear motivation seems to affect what people see. The current study will add something to the existing literature, by replicating Raymond and O’Brien (2009) whereby it can make the existing evidence stronger
and by solving the methodological shortcoming of not giving a reward during the task.
The current study will investigate whether motivation affects the early and the late stages of perception. The possible influence on early stages of perception will be researched by using an attentional blink localization task and the possible influences on late stages will be researched by trying to replicate the study of Raymond and O’Brien (2009), thus by using an attentional blink recognition task. Participants will first perform a value-learning task, which is a replication of the value-learning task used in the study of Raymond and O’Brien (2009). All faces in this task are linked to earning money, losing money or nothing. The motivational salience and valence are manipulated, because the faces have an 80, 20 or 0 percent chance to either win or lose money. In order to express this variation in salience and valence, the term expected value is used. This is the product of value and probability (Von Neumann & Morgenstern, 1944, cited in Raymond & O’Brien, 2009). Positive and negative values are used to refer to reward and punishment, which results in the following five expected values in the current study: -0.8, -0.2, 0, 0.2, and 0.8. Every trial consists of two faces and the participants should choose the face which would maximize their winnings. Thereafter, participants will perform either the attentional blink localization task or the attentional blink recognition task. No rewards are given in these tasks. In the attentional blink localization task participants will see three rapid serial visual presentations next to each other, in which figures (T1) and a face (T2) are embedded. The task will be to discriminate the texture of T1 (circles or rectangles) and to tell on which side (left or right) the face was seen. The lag between successive target presentations is either short, which creates a reduced-attention condition, or long, which creates a full-attention condition (Raymond and O’Brien, 2009). This task is used to study the influence of motivation on the early stages of perception. The attentional blink recognition task is the same as the task used in the study of Raymond and O’Brien (2009), which means that participants will see a rapid serial visual presentation in which two targets, T1
and T2, are embedded. The task will be to discriminate the texture of T1 (circles or rectangles) and then to decide if T2 was seen in the prior value-learning task (old/new decision). In this task, long and short lags will also be alternated. This task is used to study the influence of motivation on the late stages of perception. Finally, all participants have to do a rating task, in which they will rate the likability of the faces used in the current study.
It is hypothesized that motivation affects the early and the late stages of perception. It is expected that during the value-learning task, learning will increase equally for reward and punishment related pairs and that in half of the trials with neutral pairs one face is chosen and in the other half the other face. It is also expected that each face will be localized and recognized equally well. Besides, it is expected that localization and recognition will be better after the long lag than after the short lag. Moreover, it is expected that localization and recognition of faces associated with a high expected value, regardless of valence, will be more accurate, than for faces associated with a low expected value and neutral faces. In addition, is expected that familiar faces will be localized more accurately than new faces. It is also expected that recognition of faces associated with punishment or nothing will be dramatically impaired in the short lag relative to the long lag, but this is not expected for the reward-related faces. Finally, it is expected that faces associated with reward will be found more likable than new faces or faces associated with
punishment or nothing.
Method Participants
Fifty-one people participated in the current study, of which one had a non-defined gender, 13 were male and 37 were female. Participants received participation credits and/or the money they won in the value-learning task in exchange for their participation. The average age was 25,24 (SD = 9,824) and all participants were between 18 and 63 years old.
Apparatus
A Dell Optiplex 9010 computer recorded data and presented stimuli. An Aivia Osmium keyboard was used to give the answers.
Stimuli
Examples of the stimuli are shown in Figures 1, 2, and 3. T1 consisted of circles or
rectangles and T2 consisted of one of the 22 different face stimuli (11 male and 11 female) which were pictures of real neutral faces. The 154 different stimuli used as masks in the attentional blink localization task and the attentional blink recognition task were created by cutting the inner oval of the face stimuli into a set of rectangles and randomly rearranging them. There are seven different ways in which the scrambling was done, such that the seven scramble ways times the 22 different faces resulted in 154 different masks. The different scramble approaches differ with regard to how many rectangles were taken horizontally and vertically for each face. The minimum of horizontal rectangles was three and the maximum was five. The minimum of vertical rectangles was five and the maximum was eight. Thus, the minimum number of
rectangles in one way of scrambling was 15 per face and the maximum number of rectangles in one way of scrambling was 40 rectangles per face.
Figure 1. Illustration of Figure 2. Illustration of Figure 3. Illustration of
Procedure
All participants carry out the value-learning task and continue to either the attentional blink localization task or the attentional blink recognition task. The current study therefore consists of two conditions. Half of the participants are in the attentional blink localization
condition and the other half of the participants are in the attentional blink recognition condition. Informed consent. Before starting the experiment, all participants signed an informed consent paper which stated that people were well informed about the experiment due to reading the information sheet.
Value learning. All participants first read the information sheet which explained the value-learning task. This task was a choice game, where participants saw two faces next to each other with a fixation cross in between (Figure 4). The learning pairs consisted of 12 faces in total. This means that each participant saw six pairs of faces. Each face was coupled to another face and they always appeared together. All participants saw three female pairs and three male pairs. The location of the faces was randomized from trial to trial. The total number of trials was 360 and after 120 and 240 trials there was a break. Each pair was linked to winning money, losing money or neither. Winning or losing money occurred, per face, with a probability of 0.8 or 0.2. Wins and losses were always equal to 0.05 euro. Participants had to choose the face of which they thought was associated with the highest chance to win money or the lowest chance to lose money. After the participants had chosen a face (by pressing the left or the right arrow key), the screen displayed the message “WIN” in green, “LOSE” in red, or “NOTHING” in black. This was dependent on the face they chose and the associated probability which controlled the
outcome. The screen also showed the total earnings of the participant. The outcome type assigned to each face pair was counterbalanced across participants to rule out image effects. After
task or the attentional blink recognition task. An illustration of the stimuli sequence is shown in Figure 4.
Attentional blink localization task. All participants had to read an explanation sheet of the task. After that, participants had to complete five practice trials to be sure the task was well understood. Thereafter, the actual task started, which consisted of 384 trials and after 128 and 256 trials there was a break. Each trial in this task showed a rapid serial visual presentation of 24 projections, each presented for 90 milliseconds. Each projection contained three images, side by side, either all masked stimuli, or containing T1 in the middle stream or T2 in the left or the right stream (Figure 5). T1 was, randomly per trial, presented in position eight, nine, ten, eleven or twelve in the sequence. T2 then followed anywhere between position 10 and position 20. At the end of the sequence, without time pressure, participants had to press the left or the right arrow key in order to tell if T1 had a circle or rectangle pattern. After that, without time pressure, participants had to press 1 or 2 on the keyboard, in order to identify if T2 was seen right or left. No rewards were given in this task. An illustration of the stimuli sequence is shown in Figure 5.
Time Figure 4. Illustration of the stimuli
Attentional blink recognition task. Participants first had to read an explanation sheet and carry out five practice trials, in order to be sure that the task was well understood. After finishing the practice trials, participants continued to the actual task. This task consisted of 440 trials and after 110, 220 and 330 trials there was a break. Each trial in this task showed a rapid serial visual presentation of 10 projections, each presented for 100 milliseconds. Each projection contained either a masked stimulus, T1, or T2. T1 always appeared in the first position and was then followed by a masked stimulus. T2 always appeared in the third or on the ninth position, depending on the length of the lag, and was followed by a masked stimulus. After each trial participants had to push the left or right arrow key, without time pressure, in order to tell if T1 contained circles or rectangles. After that, the participants had to push the 1 or 2 on the keyboard, without time pressure, in order to identify T2 as an old (learned in the value-learning task) or a new face. No rewards were given in this task. An illustration of the stimuli sequence is shown in Figure 6.
Time Figure 5. Illustration of the stimuli
Face rating. After finishing the second task participants had to rate the likability of all the 22 faces included in the current study. A rating scale running from one (not so likable) to seven (very likable) was used.
Interview. Finally, participants had to answer the following questions: ’what do you think was the purpose of this experiment?’, ‘What was your strategy to gain as much profit as possible in the first task?’, and ‘What was your strategy to get as much answers right as possible in the second task?’.
Results
All participants were included in the analysis of the value-learning task. However, one of the participants was excluded from the analysis of the attentional blink recognition task. This was due to the fact that, in order to get an attentional blink, T1 has to be consciously perceived. One of the participants recognized the texture of T1 below chance level which means that the
attentional blink manipulation task probably could not work. Therefore, the results of that participant were excluded from the analysis of the attentional blink recognition task. This means that the results of 26 participants were included in the analysis of the localization task and the
Figure 6. Illustration of the stimuli
results of 24 participants in the analysis of the recognition task. Value Learning
Learning performance in the value-learning task is shown in figure 7. A repeated
measures ANOVA was performed in order to find out whether there was a difference between the learning levels of win and loss pairs. Neutral pairs were not included, because both options were equally likely to be chosen and neither was a better option. There were two within-subjects variables: condition (reward and punishment pairs) and time (six bins, with 60 trials per bin). The assumption of sphericity was violated. This means that the Greenhouse-Geisser degrees of
freedom will be reported. There was no significant effect of condition F(1,50) = 0.21, p = .653. Thus, learning performance was equal for reward and punishment related pairs. However, time did show a significant effect F(3.77,188.43) = 32.33, p < .001, which means that the learning performance increased over time. There was no significant interaction between condition and time F(3.53,176.35) = 1.31, p = .270.
Figure 7. Illustration of learning performance in the value-learning task. The graph presents the average probability of choosing the face linked to the optimal choice, for each of the six bins (60
0 0.2 0.4 0.6 0.8 1 1 2 3 4 5 6 P ro ba bi li ty o f O pt im al Ch o ic e Trial Bin neutral punishment reward
trials each bin) and the reward-related, punishment related and neutral face pairs. The optimal choice is in the reward-related pairs the face associated with an 80 percent chance to win money and in the punishment related pairs the face associated with a 20 percent chance to lose money. Attentional Blink Localization Task
First of all, a repeated measures ANOVA was performed with three within-subjects variables: lag (short and long), valence (reward and punishment) and expected value (high (0.8) and low (0.2)). By performing this test, it becomes clear whether lag, valence and expected value affect the accuracy of localizing T2 and if they have an interactive effect. The assumption of sphericity was violated and therefore the Greenhouse-Geisser corrected degrees of freedom will be reported. The effect of lag on accuracy of locating T2 was significant F(1,25) = 36.60, p < .001 This means that participants localized T2 more accurately after the long lag than after the short lag (Figure 8). However, the effect of valence was not significant F(1,25) = 1.59, p = .219 and the effect of expected value was not significant either F(1,25) = 0.40, p = .533. This means that the localization of T2 was not affected by valence or the expected value of the faces. There was no significant interaction between lag and valence F(1,25) = 0.01, p = .941 and also not between lag and expected value F(1,25) = 0.02, p = .880. Moreover, no significant interaction between valence and expected value was present F(1,25) = 0.94, p = .342 and also not between lag, valence and expected value F(1,25) = 0.11, p = .749. These results are shown in Figure 8.
Figure 8. Illustration of the average accurate localization scores of T2 for the different expected values, divided in the short and long lag.
However, neutral faces were not included in the previous analysis and in order to find out whether there is an effect of no expected value on the accuracy of localizing T2, another repeated measures ANOVA was performed. Two within-subject variables were included: lag (short and long) and expected value (high probability punishment (-0.8), low probability punishment (-0.2), neutral (0), low probability reward (0.2) and high probability reward (0.8)). The assumption of sphericity was violated regarding lag and regarding the interaction between lag and expected value and therefore the corrected degrees of freedom from the Greenhouse-Geisser test will be reported for those variables. An effect of lag on the accuracy of locating T2 was found F(1,25) = 35.16, p < .001, which means that participants were more accurate at locating T2 after the long lag than after the short lag (Figure 8). However, no significant effect of expected value was found F(4,100) = 0.83, p = .511, which means that expected value did not affect the accurate
localization of T2. There was also not a significant interaction between lag and expected value F(2.77,69.25) = 0.07, p =.972. These results are also shown in Figure 8.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.8 -0.2 0 0.2 0.8 A v era ge A cc ura te L o ca li z at ion o f T 2 Expected Values Short lag Long lag
In order to test whether exposure has an influence on the accurate localization of T2, a third repeated measures ANOVA was performed with two within-subjects variables: lag (short and long) and familiarity (neutral and novel faces). The assumption of sphericity was violated and therefore the corrected Greenhouse-Geisser degrees of freedom will be reported. Lag showed a significant effect F(1,25) = 27.26, p < .001, which means that people were more accurate at locating faces that were shown after the long lag than after the short lag (Figure 9). However, familiarity did not show a significant effect F(1,25) = 0.79, p = .383, and there was no significant interaction between lag and familiarity F(1,25) = 0.01, p = .913. This means that exposure was not influencing the accuracy of locating T2. These results are shown in Figure 9.
Figure 9. Illustration of the average accurate localization scores of T2 for neutral and novel faces, divided in the short and long lag.
Face exemplar control. In order to find out if the accuracy of locating T2 was affected by differences between faces, a repeated measures ANOVA was performed with two within-subject variables: lag (short and long) and face exemplar (16 faces used in the task). The assumption of sphericity was violated and therefore, the corrected Greenhouse-Geisser degrees of freedom will be reported. Lag showed a significant effect F(1,25) = 33.04, p < .001, which means that
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Neutral Novel A v era ge L o ca li z at ion s co re s o f T2 Short lag Long lag
participants were more accurate in locating T2 after the long lag instead of the short lag (Figure 10). However, face exemplar did not show a significant result F(8.13, 203.13) = 1.48, p = .166. This means that the different faces did not affect the accurate localization of T2. There was also no significant interaction between lag and face exemplar F(7.74, 193.60) = 0.73, p = .664. These results are shown in Figure 10.
Figure 10. Illustration of the mean accurate localization of T2 per face, divided in the short and long lag.
Correlation. Since the value learning went as expected, but expected value and
familiarity did not affect the accurate localization of T2, an explorative analysis was performed. This was done in order to find out if there was a relationship between learning performance and the accurate localization of T2. The total amount of money won or lost in the value-learning task was used as an indicator of learning performance. A difference score between the mean of reward and punishment related faces taken together and the mean of neutral faces was calculated. The higher the difference score, the better participants were in localizing T2. This difference score was correlated with the learning performance of participants included in the localization task. No
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 M ea n A cc ura te L o ca li z at ion o f T 2 Faces Short lag Long lag
significant correlation between learning performance and the accurate localization of T2 was found for the localization task, r = .06, p = .758. This result is shown in Figure 11.
Figure 11. Illustration of the absent correlation between learning performance and accurate localization of T2.
Attentional Blink Recognition Task
First of all, a repeated measures ANOVA was performed with three within-subject variables: lag (short and long), valence (reward and punishment) expected value (high (0.8) and low (0.2)). By performing this test, it becomes clear whether lag, valence and expected value affect the accuracy of recognizing T2 and if they have an interactive effect. The assumption of sphericity was violated and therefore the corrected degrees of freedom of Greenhouse-Geisser will be reported. Lag showed a significant effect F(1,23) = 10.10, p = .004, which means that participants were better in recognizing T2 after the long lag than after the short lag (Figure 12). However, valence did not show a significant effect F(1,23) = 0.01, p = .973 and expected value did not show a significant effect either F(1,23) = 1.84, p =.189. This means that valence and
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -2 -1 0 1 2 3 4 D if fe re n ce s co re s
Total amount of money at the end of the value-learning task, per participant included in the localization task
expected value did not affect the accurate recognition of T2. There was also no significant interaction between lag and expected value F(1,23) = 0.25, p = .622. Moreover, there was no significant interaction between valence and expected value F(1,23) = 1.06, p = .313 and also not between lag, valence and expected value F(1,23) = 0.01, p = .911. These results are shown in Figure 12.
Figure 12. Illustration of the average accurate recognition scores of T2 for the different expected values, divided in the short and long lag.
However, neutral faces were not included in the previous analysis and in order to find out whether there is an effect of no expected value on the accuracy of recognizing T2, another repeated measures ANOVA was performed. Two within-subjects variables were included: lag (short and long) and expected value (high probability punishment (-0.8), low probability punishment (-0.2), neutral (0), low probability reward (0.2) and high probability reward (0.8)). The assumption of sphericity was violated and therefore the degrees of freedom corrected by Greenhouse-Geisser are reported. Lag showed a significant effect F(1,23) = 17.87, p < .001, which means that people were better in recognizing T2 after the long lag than after the short lag (Figure 12). However, expected value did not show a significant effect F(2.73,62.85) = 0.49, p = .676, which means that the expected value of winning, losing, or keeping the same amount of money did not affect the accuracy of recognizing T2. There was also no significant interaction
0 0.2 0.4 0.6 0.8 1 -0.8 -0.2 0 0.2 0.8 A v era ge Re co gn it ion S co re s o f T 2 Expected Values Short lag Long lag
between lag and expected value F(2.65,61.02) = 0.35, p = .766. These results are also shown in Figure 12.
Finally, the most important result of Raymond and O’Brien (2009) was that neutral and punishment related faces were better recognized after the long lag than after the short lag. However, this did not apply to faces associated with reward, which means that having a high motivation to see a face overcomes the problem caused by the attentional blink. This means that reward-related faces are equally accurately recognized after both lags. In order to test whether this effect was also present in the current study, the mean recognition of reward-related faces was calculated (for the short and the long lag). Moreover, the mean recognition of neutral and
punishment related faces was also calculated. Neutral and punishment related faces were taken together, because the results in the study of Raymond and O’Brien (2009) for those faces were the same. After that, both means were used in a paired samples t-test. No significant result was found t(23) = -0.06, p = .949. This is visible in Figure 13, which shows that all faces were recognized better after the long lag than after the short lag.
Figure 13. Illustration of the mean recognition of T2 faces as a function of expected value of the T2 stimulus, divided in the short and long lag.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.8 -0.2 0 0.2 0.8 M ea n Re co gn it ion o f T 2 F ac es Expected Values Short lag Long lag
Face exemplar control. In order to find out if the accuracy of recognizing T2 was affected by differences between faces, a repeated measures ANOVA was performed with two within-subjects variables: lag (short and long) and face exemplar (16 faces used in the task). The assumption of sphericity was violated. Therefore, the corrected degrees of freedom of
Greenhouse-Geisser are reported. Lag showed a significant effect F(1,24) = 38.28, p < .001, which means that T2 was recognized more accurately after the long lag than after the short lag (Figure 14). Face exemplar also showed a significant effect F(7.00,168.02) = 4.08, p < .001, which means that the ability to recognize T2 was affected by the face that was shown.
Furthermore, there was no significant interaction between lag and face exemplar F(4.08,97.91) = 0.46, p = .772. These results are shown in Figure 14.
Figure 14. Illustration of the mean accurate recognition of T2 per face, divided in the short and long lag.
Correlation. Since the value learning went as expected, but expected value did not affect the accurate recognition of T2, an explorative analysis was performed. This was done in order to find out if there was a relationship between learning performance and the accurate recognition of T2. The total amount of money won or lost in the value-learning task was used as an indicator of
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 M ea n A cc urat e Rec o gn it ion o f T 2 Faces Short lag Long lag
learning performance. A difference score between the mean of rewarded and punished faces taken together and the mean of neutral faces was calculated. The higher the difference score, the better participants recognized T2. This difference score was correlated with the learning
performance of participants included in the recognition task. No significant correlation between learning performance and the accurate recognition of T2 was found for the recognition task, r = .20, p = .351. This result is shown in Figure 15.
Figure 15. Illustration of the absent correlation between learning performance and the ability to recognize T2.
Rating
Another manipulation check was included, in order to find out whether expected value and exposure influenced the likability of faces used in the current study. A repeated measures ANOVA was used and there was one within-subjects variable: expected value (-0.8, -0.2, 0, 0.2, 0.8, novel). The assumption of sphericity was met. No significant result was found F(5,245) = 0.58, p = .717. This means that there was no difference in likability of the faces, regardless of familiarity and expected value. Figure 16 shows this result.
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0 0.5 1 1.5 2 2.5 3 3.5 4 D if fe re n ce S co re s
Total amount of money at the end of the value-learning task, per participant included in the recognition task
Figure 16. Illustration of the mean rating scores of neutral faces, novel faces and faces associated with the different expected values.
Discussion
The current study investigated if motivation affects the early and the late stages of perception, by looking at whether reward-related information was prioritized in perception. The early stages of perception are the stages in which detection of stimuli takes place and the late stages are the stages in which recognition and discrimination of stimuli takes place. All
participants first had to perform a value-learning task. All faces used in this task were linked to winning money, losing money or nothing. The motivational salience and valence were
manipulated, because the faces had an 80, 20 or 0 percent chance to either win or lose money. After that, participants continued to either the attentional blink localization task or the attentional blink recognition task. The localization task can be seen as the one researching the early stages of perception and the recognition task as the one researching the late stages of perception. No rewards were given in these tasks. Finally, all participants had to do a rating task, in which they rated the likability of all the faces used in the current study.
Prior to the study was hypothesized that motivation would affect the early and the late stages of perception. Moreover, it was expected that during the value-learning task, learning
0 1 2 3 4 5 -0,8 -0,2 0 0,2 0,8 novel M ea n Ra ti n g S co re Expected Values
would increase equally for reward and punishment related pairs and that in half of the trials with neutral pairs, one face is chosen and in the other half the other face. It was also expected that each face would be localized and recognized equally well. Furthermore, it was expected that faces would be recognized and localized more accurately after a long lag than after a short lag. Moreover, it was expected that localization and recognition of faces would be more accurate when faces were associated with a high expected value, regardless of valence, instead of a low expected value or nothing. It was also expected that familiar faces would be localized more accurately than new faces. In addition, it was expected that recognition of faces associated with punishment or nothing would be dramatically impaired in the short-lag relative to the long lag, but this was not expected for the reward-related faces. Finally, it was expected that faces associated with reward would be found more likable than new faces or faces associated with punishment or nothing.
The results showed that learning performance for win and loss pairs was equal. Moreover, the results showed that localization of T2 was better after the long lag than after the short lag. However, valence, expected value and exposure did not affect the accurate localization of T2. This goes partly against the expectations set prior to the study. Besides, the results also showed that all faces were localized equally well and that there was no correlation between learning performance and accuracy of locating T2. The results also showed that T2 was recognized more accurately after the long lag instead of the short lag. However, valence and expected value did not affect the accuracy of recognizing T2. This also goes partly against the expectations set prior to the study. Moreover, the results showed that recognition of faces associated with punishment, reward or nothing was better after the long lag than after the short lag and there was no difference in the accurate recognition of faces associated with reward and faces associated with punishment or nothing. This goes against the expectations set prior to the study as well. In addition, the
results showed a difference in recognizing T2 dependent on face exemplars. This also goes against the expectations set prior to the study. Moreover, the results also showed that there was no correlation between learning performance and accurate recognition of T2. Finally, the results showed that there was no difference in likability of the faces, thus expected value and exposure did not affect the likability of the faces. This goes also against the expectations set prior to the study. On the basis of the results of this study, it appears that motivation does not affect the early and late stages of perception. This means that this study does not provide evidence for influences of top-down effects on perception.
The results of the current study are partly in line with previous studies classified as, in this paper, studies that examined early stages of perception. An earlier study, done by Milders,
Sahraie, Logan and Donellon (2006) showed that motivation, operationalized by fear
conditioning, can modulate detection of formerly neutral stimuli earlier. In addition, Anderson, Siegel, Bliss-Moreau and Feldman Barrett (2011) showed that motivation, operationalized as negative, positive or non-social information, affected the dominance duration of faces paired with descriptions of negative, positive or non-social behaviour. The current study is not in line with the just mentioned studies, because it was not able to show the influence of motivation on early stages of perception. However, these studies contained some methodological shortcomings. The current study solved these shortcomings, by using neutral stimuli and a performance-based task. However, previous studies tried to replicate the study of Anderson et al. (2011), but were not able to find the same results (Rabovsky, Stein & Abdel Rahman, 2016; Stein, Grubb, Bertrand, Suh & Verosky, 2017). These studies were unable to find influences of motivation, operationalized as affective knowledge, on suppression durations of faces. The current study is in line with the just mentioned studies, because this study was also not able to show an influence of motivation on the early stages of perception.
However, the results of the current study are not in line with previous studies classified as, in this paper, studies that examined the late stages of perception. An earlier done study showed that perception is affected by reward and punishment in opposite directions (Wilbertz, Van Slooten & Sterzer, 2014). However, this study contains a methodological shortcoming. The current study has overcome this problem, by learning the motivation before starting the attentional blink tasks. Moreover, Müsch, Engel and Schneider (2012) showed that there is an attentional advantage for emotional faces and easy to distinguish stimuli. In addition, Raymond and O’Brien (2009) showed that regardless of available attention, recognition was substantially enhanced for motivationally salient stimuli, compared to equally familiar stimuli that had weak or no motivational salience. This effect was found regardless of valence. The current study is not in line with previous studies, because this study was unable to replicate the study of Raymond and O’Brien (2009) and to show any motivational effect on recognition.
Thus, the current study is unable to confirm what earlier studies suggested, namely that there might be influences of top-down effects on early stages of perception and that there are certainly influences of top-down effects on late stages of perception. However, the current study was the first one making a distinction between different kinds of perception, whereby it can become clearer which part of perception is affected by top-down effects. However, based on this study motivation does not seem to affect the early and the late stages of perception.
A possible alternative explanation for the unexpected result of not being able to replicate the study of Raymond and O’Brien (2009) is that this study is not an exact replication. However, in psychology it is not possible to perform an exact replication (Rosenthal, 1991, cited in Brandt et al., 2014) and therefore it is important to stay as close as possible to the original study. A convincing close replication follows, among other things, the rule of following as exactly as possible the methods of the original study, of which stimuli the are part (Brandt et al., 2014).
However, the current study used pictures of real neutral faces instead of the computer-generated faces that Raymond and O’Brien (2009) used. The results showed that there was a difference in the accuracy of recognizing faces. This means that some faces were recognized better than others, regardless of the length of the lag. This creates the possibility that recognition of T2 was not influenced by motivation, but by the characteristics of certain faces. Thus, follow-up studies should use neutral computer-generated faces in order to rule out the possible effects of characteristics of individual faces.
A shortcoming of this study is that some participants declared in the interview that they only focused on the left or the right stream of images in the attentional blink localization task, even though they were instructed to focus on the middle stream.It is then possible to locate T2 by only determining whether or not it was on the side that was focused on. This means that to give the right answer, it is not necessary to have seen T2. Thus, this task might not have tested the actual ability to localize T2, but instead tested if participants saw something in one stream or not. Follow-up studies could include a task that is set up in such a way that actual localization must take place. This could be done by, for example, showing a rapid serial visual presentation in the middle in which T1 will be shown and showing rapid serial visual presentations above, below and on the sides of the middle stream in which T2 will be shown. Then it would be impossible to focus only on one stream to find out where T2 was shown. This would be a better
operationalization of localization, whereby the found results of the current study might be different in a follow-up study.
Another shortcoming is that there is a small group of participants in the current study. While doing research, it is important to aim for a power of 0.8. This means that there will be an 80% chance to detect an effect if one genuinely exists (Field, 2014). This study contained 51 participants in the value-learning task and in the rating task, which is related to a power of 0.94
(94%). However, only 26 participants were included in the localization task, which is related to a power of 0.69 (69%). Moreover, 24 participants were included in the recognition task, which is related to a power of 0.65 (65%). Thus, the chances of detecting an effect were for both the localization task as the recognition task low. Having a small group of participants also means that the found results might not account for the entire population, because the higher the number of participants, the higher the chance of getting reliable results that account for the entire population (Mook & Parker, 2001). Further research can solve this shortcoming by including more
participants.
In conclusion, it can be stated that this study is not able to support the idea that motivation affects the early and the late stages of perception. The results showed that learning performance was equal for faces associated with reward and faces associated with punishment. Moreover, localization and recognition of T2 was more accurate after the long lag instead of the short lag. However, no effect of valence and expected value on localization and recognition of T2 were found. Moreover, no effect of exposure on the accurate localization of T2 was found. There were also no correlations found between learning performance and accuracy of localizing and
recognizing T2. In addition, an effect of face exemplar on the accuracy of recognizing T2 was found, which was not found for localizing T2. The most striking result was that recognition or faces associated with punishment, reward or nothing was better after a long instead of short lag. Herewith was no difference between faces associated with reward and faces associated with punishment or nothing, which is in contrast to what Raymond and O’Brien (2009) found. This unexpected result could be due to the fact that some faces in the recognition task were better recognized than others. Therefore, neutral computer-generated faces should be used in further research. Moreover, further research should obviate the methodological shortcomings, the
is not entirely in line with earlier studies and that this study is not able to replicate the study of Raymond and O’Brien (2009), more research should be done to show whether top-down effects, such as motivation, affect the early and the late stages of perception. The answer to this question remains important in vision science, because of the possibility that outdated models are used.
Literature list
Anderson, E., Siegel, E. H., Bliss-Moreau, E., & Feldman Barrett, L. (2011). The visual impact of gossip. Science, 332, 1446-1448.
Brandt, M. J., Ijzerman, H., Dijksterhuis, A., Farach, F. J., Geller, J., Giner-Sorolla, R., Grange, J. A., Perugini, M., Spies, J. R., & Veer, A. van ‘t (2013). The replication recipe: What makes for a convincing replication? Journal of Experimental Social Psychology, 50, 217-224.
Bruner, J. S., & Goodman, C. C. (1947). Value and need as organizing factors in perception. The Journal of Abnormal and Social Psychology, 42, 33-44.
Field, A. (2014). Discovering statistics using IBM SPSS statistics. London: SAGE.
Firestone, C., & Scholl, B. J. (2016). Cognition does not affect perception: Evaluating the evidence for “top-down” effects. Behavioral and Brain Sciences, 39, 1-77.
Milders, M., Sahraie, A., Logan, S., & Donnellon, N. (2006). Awareness of faces is modulated by their emotional meaning. Emotion, 6, 10-17.
Mook, D. G., & Parker, S. (2001). Psychological Research: The Ideas Behind the Methods. New York: W.W. Norton and Company.
Müsch, K., Engel, A. K., & Schneider, T. R. (2012). On the blink: The importance of target- distractor similarity in eliciting an attentional blink with faces. PLoS ONE, 7, 7, 1-10.
Rabovsky, M., Stein, T., & Abdel Rahman, R. (2016). Access to awareness for faces during continuous flash suppression is not modulated by affective knowledge. PLoS ONE, 11, 4, 1-17.
Raymond, J. E., & O’Brien, J. L. (2009). Selective visual attention and motivation: The consequences of value learning in an attentional blink task. Psychological Science, 20, 981-988.
Shapiro, K. L., Raymond, J.E., & Arnell, K. M. (1997). The attentional blink. Trends in Cognitive Sciences, 1, 291-296.
Stein, T., Grubb, C., Bertrand, M., Suh, S. M., & Verosky, S. C. (2017). No impact of affective person knowledge on visual awareness: Evidence from binocular rivalry and continuous flash suppression. Emotion, 1-9.
Tsuchiya, N., & Koch, C. (2005). Continuous flash suppression reduces negative afterimages. Nature Neuroscience, 8, 1096-1101.
Wilbertz, G., Slooten, J. van, & Sterzer, P. (2014). Reinforcement of perceptual inference: Reward and punishment alter conscious visual perception during binocular rivalry. Frontiers in Psychology, 5, 1-9.
