the aha!” moment: alaina baker submitted to the department ... · the a-ha! moment 2 abstract we...

16
The A-ha! Moment 1 The "Aha!” moment: How prior knowledge helps disambiguate ambiguous information Alaina Baker Submitted to the Department of Psychology of Northeastern University for the degree of Bachelor of Science in Psychology with Honors in the Discipline Lisa Feldman Barrett, PhD, Honors Project Faculty Advisor April, 2017

Upload: others

Post on 17-Sep-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 1

The "Aha!” moment:

How prior knowledge helps disambiguate ambiguous information

Alaina Baker

Submitted to the Department of Psychology

of Northeastern University

for

the degree of Bachelor of Science in Psychology

with Honors in the Discipline

Lisa Feldman Barrett, PhD, Honors Project Faculty Advisor

April, 2017

Page 2: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 2

Abstract

We encounter ambiguous information every day. Previous research suggests that prior

knowledge is necessary for making sense of this information and that such meaning-making is

effortless and automatic (Barrett, 2017; Barrett & Bar, 2009). The present study sought to

confirm and extend this idea in the visual domain. We hypothesized that participants would be

able to disambiguate ambiguous visual information more easily and quickly after exposure to

relevant perceptual knowledge. Moreover, we predicted that older participants would have

greater success in disambiguating information, even before presentation of relevant perceptual

information, given their greater exposure to relevant perceptual experiences over their lifetime.

We recruited 68 participants at the Museum of Science, Boston, who rated their ability to see

objects in a series of ambiguous (distorted) images both before and after being exposed to a clear

(non-distorted) version of the image. In line with our hypotheses, participants rated the

ambiguous image as less ambiguous after exposure to the corresponding clear (original) image

and did so more quickly. Further, we found that neutral images were more easily and quickly

disambiguated than positive or negative images. Overall, these findings reveal the significance of

previous experience in alleviating our “experiential blindness” and making sense of the

perceptual ambiguities in our world.

Page 3: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 3

The "Aha!” moment:

How prior knowledge helps disambiguate ambiguous information

If you were presented with the photo below (Figure 1) and asked to make sense of its

contents, could you? If you turn to appendix A, you will see a clearer image; once you look at it,

return to this ambiguous one.

Figure 1: Ambiguous Image

Now that you have relevant perceptual experience with the content of this photo, it is easier to

make sense of its once entirely ambiguous contents. Moreover, this change in your perception

probably occurred effortlessly and automatically. Without the previous experience and

knowledge, though, you likely could not disambiguate the image. That is, you were

“experientially blind”. This demonstration illustrates an important feature of perception. We tend

to believe that our visual experiences are driven entirely by the external world: we see an object

how it ‘really is’ because we are just passively taking in wavelengths of light. However,

emerging neuroscience evidence suggests that, not only do internal contexts (like how we feel

and our past experience) influence perception, but they actually drive perception (Barrett, 2017;

Page 4: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 4

Barrett & Bar, 2009). The brain’s prediction for what it will see in the next moment shapes

perceptual experience before it occurs.

Perception involves the integration of sensory input and prior knowledge (e.g.,

Summerfield & de Lange, 2014). Contrary to more classical views of perception, predictive

coding theories of perception (e.g., Barrett, 2017; Clark, 2013) posit that prior knowledge

(predictions) actually precede and actively shape the processing of incoming sensory information

in real time. That is, the brain is constantly attempting to match incoming sensory inputs with

apriori expectations or predictions. The brain attempts to “explain away” sensory input by

making its best guess about what it will see (or hear, or feel, etc.) in the next moment.

Components of the sensory signal that coincide with (i.e., are predicted by) the current “winning

hypothesis” are not processed further; the perception of these components becomes what was

predicted. Unexplained (i.e., not predicted) components of the sensory signal are transmitted up

the predictive hierarchy as prediction error. The better the match, the less prediction error that

climbs that predictive hierarchy. Any prediction error that does flow up the hierarchy can modify

the internal model that the brain is using to generate the predictions, leading to better predictions

in the future (Barrett, 2017; Clark, 2013; Summerfield & de Lange, 2014). In this way,

unpredicted sensory information can inform future predictions, helping the brain improve its

sensory predictions for similar stimuli or experiences in the future.

According to these theories, prior knowledge will influence perception the most when

incoming sensory information is ambiguous or imprecise (Summerfield & de Lange, 2014).

Given that predictions are so critical in the process of perception and that predictions are shaped

by your prior experiences and knowledge, structural regularities in visual information allow the

formation of more accurate expectations about future sensory stimulation. Thus, repeated

Page 5: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 5

sensory input normally reduces the corresponding neuronal responses because it subsequently

increases predictability (Clark, 2013). Therefore, the more you experience and learn from new

experiences, the better your predictions and the more efficient your perception should become.

The current experiment aims to further investigate how “experiential blindness” is

resolved; that is, how ambiguous information is disambiguated through the use of prior

knowledge. We examine this by asking participants to identify ambiguous images before and

after exposure to a clearer version of the same image (called the original image). Using this

paradigm, we will test the hypothesis that prior knowledge aids in disambiguating ambiguous

perceptual information. We make two specific predictions related to this hypothesis. First, we

predict that participants will be able to disambiguate ambiguous images more easily and quickly

after exposure to the unambiguous, original version of those images (i.e., after they have prior

knowledge or perceptual experience on which to draw). Second, we predict that disambiguation

ability should increase with age. Because older individuals will likely have been exposed to a

larger variety of objects and situations throughout their longer lifetimes, they should have a

larger variety of prior experience from which they can build predictions to use in disambiguating

the world around them. Thus, in the present study, we predict that older participants will have

greater success in disambiguating the images (due to their greater prior experience), even before

presentation of the unambiguous, original image. Finally, as an exploratory analysis we examine

whether the affective valence of the images (i.e., neutral, positive, or negative) influences how

prior knowledge is utilized to disambiguate ambiguous perceptual information.

Page 6: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 6

Method

Participants

The sample consisted of 68 visitors (34 male, 34 female) to the Museum of Science in

Boston, MA, who voluntarily chose to engage with the researcher while at the Museum. The

final sample ranged from age 8 to age 77 (M=28.15 years, SD=16.27 years) and was comprised

of 77% White, 7% Black, and 7% Asian participants. Nine percent of participants identified

themselves as more than one race. To be eligible, potential participants needed to be at least six

years of age, have normal or corrected-to-normal vision, and speak English. Participants

completed the experiment in one experimental session, lasting 10-15 min, with the researcher in

the Hall of Human Life at the Museum of Science. Participants were not compensated

monetarily, but the experiment was “advertised” at the museum as a chance to “learn about how

scientists investigate a wide array of topics related to human biology and health and help

advance these fields through [their] participation.”

Materials

This study utilized a set of 68 original (clear) images of varying objects (e.g., animals,

plants, foods), and a set of 68 ambiguous versions of these images (one matched to each of the

clear images; See Figure 2). To create the ambiguous version of each image, a clear (original)

version of an image was imported into Gimp software. There, the image mode was switched to

grayscale, the colors were inverted, and the ‘artistic oilify’ filter, which uses a very high-contrast

filter, was added to degrade the clarity of the image. Images were selected from the internet such

that they fit into one of three categories based on their affective valence at face-value; images

were positive (e.g., an image of a kitten), neutral (e.g., an image of clothes), or negative (e.g., an

image of a snarling cheetah). The final set of 68 paired images were selected from a larger set of

Page 7: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 7

262 images that were normed for valence, arousal, and ambiguity on Amazon’s Mechanical Turk

(N=100). Ratings of pleasantness (e.g., this image made me feel intensely unpleasant or intensely

pleasant) and ratings of activation (e.g., this image made me feel intensely deactivated or

intensely activated) after viewing each image confirmed that images fit the pre-chosen affective

categories to which they were assigned (i.e., positive, neutral, negative). The 68 images utilized

in the current study were chosen with the maximal combination of (1) very high ratings of

ambiguity for the first image and (2) very low ratings of ambiguity for the second image (3) in

their respective affective categories (neutral, positive, negative).

Figure 2: Ambiguous (left) and Original (right) Image Pair

Procedure

Consent and Assent Process: Depending on the age of the participant, a consent form was

signed. If the participant was over 18, verbal consent was given after thoroughly reviewing the

form with the researcher. If the participant was under 18, a consent form was reviewed and

signed by a parent and/or guardian. If the participant was a child who was unable to read, an

assent form was read to the participant by a researcher.

Page 8: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 8

Ambiguous Image Task: This task was conducted on a laptop and was run using E-Prime

software (version 2.0; Psychology Software Tools, INC.). On each trial of this Ambiguous Image

Task, an ambiguous image was presented for 4 seconds. Participants were then asked, “Did you

see anything?” which they rated on a four-point scale either “definitely yes” (1), “a little bit” (2),

“not really” (3), or “definitely no” (4). If they rated (1) or (2) they were asked a second question,

“what did you see?”, which they answered aloud. The researcher recorded the participant’s

answer(s) in a separate spreadsheet. The participant was then shown the original image for 2

seconds, followed by its matched ambiguous image a second time for 2 seconds, after which the

participants answered the same two questions again. Participants each completed 26 trials of this

task, where the 26 original images and matched ambiguous images they saw were drawn at

random from the total set of 68 available pairs of images.

Post-Experimental Questionnaire and Demographic Information Survey: Following the

Ambiguous Image Task, participants answered a brief questionnaire about their experience(s)

during the task (see Appendix B for experimental questionnaire) and completed a demographic

survey where they were asked to report their gender, age, race, and ethnicity. If participants

were unlikely to know their own demographic information and/or were under 18 years of age,

the demographic form was given to the parent/guardian to fill out for the particpant.

Debriefing: Lastly, participants were guided through a debriefing form by the researcher

and were given the opportunity to ask questions about the experiment. They were also given two

stickers (one provided by the museum and one provided by the researcher) to indicate their

participation.

Page 9: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 9

Results

Ratings of Ambiguity. A 3x2 repeated-measures ANOVA, with valence of the image

(positive vs. negative vs. neutral) and presentation number (first presentation vs. second

presentation) as within-subjects measures, revealed a significant main effect of valence on

ratings of ambiguity, F(2, 65)= 23.05, p<.001. A post-hoc Fisher’s least significant difference

test revealed that neutral images were rated as significantly less ambiguous (M=2.25, SE=.05)

than both positive (M=2.55, SE=.07) and negative (M=2.70, SE=.08) images, ps<.001, which did

not differ in rated ambiguity, p=.06. Consistent with predictions, this analysis also revealed a

main effect for presentation number (F(1, 65)= 104.98, p<.001), such that the ambiguous images

were rated as more perceptually ambiguous following their first presentation (M=2.79, SE=.06)

than their second presentation (M=2.21, SE=.07). Results did not reveal a significant interaction

between valence and presentation number, F(2, 65)=2.44, p=.09, suggesting that the effect of

image valence was consistent across both presentations of the ambiguous image. See Figure 3.

Figure 3. Ratings of perceptual ambiguity by image valence and presentation number.

1.8

2

2.2

2.4

2.6

2.8

3

3.2

First Presentation Second Presentation

Self-

repo

rted

Perc

eptu

al

Am

bigu

ity

Neutral

Positive

Negative

Page 10: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 10

Reaction Times. A similar 3x2 repeated-measures ANOVA failed to reveal either a

significant main effect of valence on reaction times, F(2, 65)= 1.50, p=.23, or a significant

interaction between valence and presentation number on reaction times, F(2, 65)=1.24, p=.29.

However, as predicted, this analysis did reveal a significant main effect for presentation number,

F(1, 65)= 20.51, p<.001, such that the ambiguous images were rated more slowly after the first

presentation (M=3635.16, SE=258.87) than the second presentation (M=2848.62, SE=249.62) of

the ambiguous images. See Figure 4.

Figure 4. Ratings of reaction time by image valence and presentation number.

Age-related Effects. Contrary to predictions, there were no significant correlations

between participants’ ages and their ratings of perceptual ambiguity for the ambiguous image on

either its first presentation, r(67)=0.17, p=0.18, or its second presentation, r(67)=.08, p=.53.

Discussion

In support of our hypothesis, results demonstrated that participants rated ambiguity of the

image more quickly (i.e., shorter reaction time) following the second presentation of the

ambiguous image than the first presentation. This supports the idea that disambiguation of

0

500

1000

1500

2000

2500

3000

3500

4000

4500

First Presentation Second Presentation

Rea

ctio

n Ti

me

(ms)

Neutral

Positive

Negative

Page 11: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 11

ambiguous percepts occurs faster when a person has had prior exposure to a similar (or identical)

percept. Prior knowledge might help guide participants’ attention to salient aspects of the

ambiguous display, leading to its perceptual resolution. Also in line with our hypothesis, results

additionally demonstrated that the ability to disambiguate an image significantly increased from

the first presentation of the ambiguous image to the second presentation of the ambiguous image

following exposure to an original, unambiguous version of the image. This suggests that

participants can more successfully form coherent percepts when they have prior perceptual

experience or knowledge from which to draw.

Contrary to our hypothesis, results did not demonstrate that perceptual ability increased

with age. We assumed age might be associated with greater prior experience and knowledge,

allowing older individuals to resolve perceptual ambiguity more aptly and more quickly. Our

findings suggest that perhaps age is not the best indicator of previous perceptual experience.

Instead, future research might focus on other variables that might better approximate the amount

of past relevant experience an individual has to draw from, such as expertise in a specific domain

(e.g., knowledge of animals or food). For example, if a 6-year-old child happens to be an expert

on animal species, he or she might be able to resolve perceptual ambiguity related to animals

more quickly and accurately than an elderly individual who has seen few of the same stimuli in

his or her lifetime. However, the present study may also have been insufficiently powered to

detect this kind of effect, despite recruiting participants with a very wide range of ages, so future

studies should recruit larger samples of relevant age groups.

Further, exploratory analyses revealed that image valence impacted ratings of perceptual

ambiguity at both presentations of the ambiguous image (both before and after exposure to the

unambiguous original image). For example, whether an image was considered neutral (e.g., a

Page 12: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 12

mushroom), positive (e.g., a kitten), or negative (e.g., a bear) influenced an individual’s ability to

disambiguate the ambiguous images, such that neutral images were rated as the least ambiguous

when compared to positive or negative images. This could be due to a variety of factors. First,

the experiment was conducted at the Museum of Science, where strict constraints were placed on

the emotional evocativeness of the stimuli allowed to be utilized in the study. Because of this

limitation, participants saw very few negative images and many more neutral images within the

Ambiguous Image Task. In a way, this may have trained participants to expect to encounter

certain kinds of images within the task itself (i.e., neutral images) over others (i.e., negative

images). That is, participants may have been using their prior experience within the task to shape

their perceptual predictions about what they would likely see on the next trial (i.e., another

neutral image), and predicted images should be more ably and quickly disambiguated. A second

explanation of this finding is that participants may have more prior experience with the objects in

the neutral images than those in the more evocative images because of the familiarity of and

constant exposure to the (neutral) everyday objects depicted in these images; things that are

predicted should, in fact, appear less ambiguous. Finally, a third possible explanation is that this

finding may have also been due to uncontrolled low-level visual features of the images from the

different valence categories, where the neutral images may have been easier to disambiguate, not

because of their familiarity, predictability, or valence, but because of visual properties inherent to

the images themselves.

Future research should further explore the influence of image valence on perceptual

ambiguity. In the current study, highly negative images (e.g., a photo of a venomous snake)

could not be used. Future research could include more trials of non-neutral images, particularly

those with negative valence. By including these images, we might better reveal the impact of

Page 13: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 13

emotional salience on how prior knowledge is deployed to make sense of ambiguous

information. Future research could also look at individual differences that might moderate these

factors for particular images. For example, is someone who has had a frightening encounter with

snakes more likely to disambiguate a negative image of a snake faster and more accurately? It is

possible he or she might disambiguate the image faster, but only when a snake is present. He or

she should be less accurate when the object has some snake-like features, but is not a snake.

Similarly, if someone has an affective disorder marked by increased negative affect (e.g., major

depressive disorder), will he or she be more likely to see less perceptual ambiguity when viewing

negatively-valenced images? Indeed, a recent study by Teufel et al. (2015), which used similar

ambiguous image stimuli, showed that psychosis was related to greater reliance on prior

knowledge and hence better disambiguation ability.

The current findings demonstrate the importance of previous experience and knowledge

in perceiving ambiguous stimuli. Though we often believe that we perceive and experience

things as they happen in the immediate environment, the results of this experiment add to the

growing body of evidence demonstrating that internal contexts drive perception. Our brain is

constantly issuing predictions that alter what we see in the next moment, and we are able to

correct what we fail to predict accurately in one moment because we are constantly accruing new

experiences and updating our predictions every moment of every day.

Page 14: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 14

References:

Barrett, L. F. (2017). How emotions are made: The secret life of the brain. Houghton Mifflin

Harcourt.

Barrett, L. F., & Bar, M. (2009). See it with feeling: Affective predictions during object

perception. Philosophical Transactions of the Royal Society B: Biological Sciences,

364(1521), 1325-1334.

Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive

science. Behavioral and Brain Sciences, 36(03), 181-204.

Summerfield, C., & De Lange, F. P. (2014). Expectation in perceptual decision making: Neural

and computational mechanisms. Nature Reviews Neuroscience, 15(11), 745-756.

Teufel, C., Subramaniam, N., Dobler, V., Perez, J., Finnemann, J., Mehta, P. R., ... & Fletcher, P.

C. (2015). Shift toward prior knowledge confers a perceptual advantage in early

psychosis and psychosis-prone healthy individuals. Proceedings of the National Academy

of Sciences, 112(43), 13401-13406.

Page 15: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 15

Appendix A. Original Image Version of Figure 1

Page 16: The Aha!” moment: Alaina Baker Submitted to the Department ... · The A-ha! Moment 2 Abstract We encounter ambiguous information every day. Previous research suggests that prior

The A-ha! Moment 16

Appendix B. Experimental Questionnaire