Pre-existing categorization diminishes attractive and repulsive temporal history effects on perception

class: left, middle, inverse, title-slide

.title[
# <span style="font-size:58px">Pre-existing categorization diminishes attractive and repulsive temporal history effects on perception</span>
]
.author[
### Eline Van Geert<sup>1</sup>, & Johan Wagemans<sup>1</sup>
]
.institute[
### <sup>1</sup>KU Leuven, Belgium
]
.date[
### ICPS Brussels, Symposium Categorization across Fields: Insights from Learning, Cognition, and Perception </br> March 9<sup>th</sup>, 2023
]

---

layout: true

.footer_right[]
.footer_middle[<a href="https://tinyurl.com/ICPSBRU23" target="_blank">slides at tinyurl.com/ICPSBRU23</a>]

---
class: top, normal, center
### Perception depends on...

.footnote[<a href = "https://doi.org/10.3389/fnhum.2015.00594"> Snyder et al. (2015)</a>; <a href = "https://doi.org/10.1177/20416695221109300"> Van Geert, Moors, Haaf, & Wagemans (2022)</a></br></br>]

???

Hello everyone, Today I will present you my work on the relation between categorization and immediate temporal context effects on perception.

From previous research we know that perception does not only depend on the stimulus that is shown on a certain point in time, but also depends on what was shown and perceived just before the current stimulus was presented.

Earlier research found two different temporal context effects.

First, when we consider the effect of the previous percept on the current percept, we speak of hysteresis. This refers to an attractive effect of the previous percept, or in other words, a tendency to organize the visual input in a similar way as preceding context stimuli. When we perceive an ambiguous stimulus at time 1 as a car, the chance to perceive a car again at the second timepoint will increase.

---
class: top, normal, center
### Perception depends on...

<center>... what was just <b>shown</b></center>
</div>

???
Second, when we consider the effect of the previous stimulus on the current percept, we speak of "adaptation". Adaptation refers to a repulsive effect of the previous stimulus evidence, or in other words, a tendency that pushes the current percept away from the organization that was most dominant in the previous stimulus. When we are presented with a very clear example of a car at time 1 and perceive it as a car, the chance to perceive a car again at the second timepoint will decrease compared to when a less clear example of a car was presented at time 1.

---
class: top, normal

### Data collection

.pull-left[
- 283 first-year psychology students (88% female)

- 3 recognizable and 3 non-recognizable morph series

]

???

In the study that I will present to you today, 283 first year psychology students participated in a categorization task, a discrimination task, and a similarity judgment task. 
Each participant completed each task for a different recognizable and non-recognizable series

---
class: top, normal

### Data collection

.pull-left[

- 283 first-year psychology students (88% female)

- 3 recognizable and 3 non-recognizable morph series

<img src="figs/line_set1b.png" width="95%" /><img src="figs/line_set3b.png" width="95%" /><img src="figs/line_set4a.png" width="95%" />
]

.pull-right[
<img src="figs/cattask.png" width="100%" />

]

???

In this presentation I will focus on the results for the categorization task (for which you see the trial structure presented here). Before starting the categorization task, participants were presented with 4 examples of category A and 4 examples of category B, that is the only information they received. They did not receive any form of feedback during the task.

---
class: top, normal

### Data collection

.pull-left[

- 283 first-year psychology students (88% female)

- 3 recognizable and 3 non-recognizable morph series

]

.pull-right[
<img src="figs/cattask.png" width="100%" />

</br>

]

???

We will indicate the stimulus evidence or morph level with a number from -5 to 5, with -5 being the stimulus that provides most evidence for category A, 5 being the stimulus that provides most evidence for category B, and 0 being the stimulus exactly in between category A and B.

---
class: top, normal

### Categorization responses

.footnote[<a href = "https://doi.org/10.31234/osf.io/6x75c"> Van Geert & Wagemans (2022)</a></br></br>]

???
First I show you the overall categorization responses.

On the x axis: morph level
On the y axis: probability of perceiving the current stimulus as part of category B

the top row of graphs shows the results for the recognizable morph series
the bottom row of graphs shows the results for the non-recognizable morph series

stronger categorization for the recognizable morph series compared to the non-recognizable morph series

What about the temporal context effects (effects previous percept and previous stimulus)?

---
class: top, normal

### Attractive context effect previously perceived category

???

For the non-recognizable morph series, we see a clear, overall, attractive effect of the previously perceived category, a clear hysteresis effect.
The chance of perceiving the current stimulus as category B is higher when you have just perceived the previous stimulus B, compared to having perceived the previous stimulus as A.

This attractive effect of the previously perceived category is largely absent in the recognizable morph series.

---
class: top, normal

### Attractive context effect previous perceived category

???

Here we see the same results but averaged across the different recognizable and non-recognizable morph series.

---
class: top, normal

### Attractive context effect diminishes over time

???
When we split the data for the different blocks part of the experiment, we see that the attractive effect of the previously perceived category present in the data for the non-recognizable morph series diminishes over time. This result is in line with the idea that stronger categorization diminishes the effect of the immediate context on perception.

---
class: top, normal

### Repulsive context effect previous category evidence

???

Let's first have a look at the results for the non-recognizable morph series.

On the x axis: current morph level
On the y axis: probability of perceiving the current stimulus as category B
In the panels: previous percept / previously perceived category indicated
In the legend (colors): previous stimulus/category evidence

The effect of the previous stimulus evidence on the current percept is dependent on the congruency between the previous percept and the current stimulus.

When the previous stimulus was perceived as part of category A, and the current stimulus is in favor of category B, you see a typical repulsive adaptation effect: the chance of perceiving the current stimulus as B is then higher when the previous stimulus was in favor of A (than when the previous stimulus was in favor of B). As we can see in the rectangle in the left plot, the light lines lay higher than the darker lines.

Similarly, in the rectangle in the right plot, when the previous stimulus was perceived as B and the current stimulus is in favor of A, the chance of perceiving the current stimulus as A is higher when the previous stimulus was in favor of B.

However, when the current stimulus is in favor of the same category as the previous percept, we do not see a lot of evidence for such a repulsive adaptation effect.

This effect is clear in the non-recognizable morph series, but absent in the recognizable morph series.
In the recognizable morph series, this effect of the previous stimulus is not evident.
We could thus conclude that these context effects of previous category evidence are present only in the  non-recognizable morph series
→ in other words, one could say that stronger categorization diminishes the use of immediate context information in forming the current percept.

[because less data points per condition per series, here combined across different morph series]

---
class: top, normal

### Repulsive context effect previous category evidence

???

[because less data points per condition per series, here combined across different morph series]

---
class: top, normal

### No indication for temporal change in repulsive context effect

???

When we plot the results for the different blocks, we do not see any indications for a temporal change in the size of the repulsive effect of the previous stimulus evidence.

---
class: top, normal

### Conclusion

**For non-recognizable morph series:**

- clear **attractive effect** of **previously perceived category** (i.e., hysteresis)

- when current stimulus incongruent with previous percept, **repulsive effect** of **previous category evidence** (i.e., adaptation)

</br></br>

**No context effects present for recognizable morph series** (i.e., when strong categorization is present)

???

We found a clear attractive effect of the previously perceived category: a clear hysteresis effect

When the current stimulus was from a different category than the previous percept, there was a repulsive effect of the previous stimulus evidence present.

Important to distinguish between the effects of the previous stimulus and the previous percept: both effects need to be taken into account to understand the pattern of results present in this study on shape categorization.

We also replicated these findings for two of the non-recognizable morph series in a separate study

---
class: top, normal

### Takeaway

</br>

<b>&rarr; Strong pre-existing categorization diminishes attractive and repulsive temporal history effects on perception</b>

</div>

???
As a key takeaway, this study brings evidence for the idea that strong pre-existing categorization can diminish attractive and repulsive temporal history effects on perception. That is, in case a clear categorization is present, we will use this categorization to clarify our percept. In case no clear categorization is present, we make more use of immediate temporal context to clarify our percept.

---
class: top, normal

<center>
<img src = "img/EVG_OCTA.svg" style = "width: 50%;" margin-bottom: -10px;">
<div style = "margin-top: -30px;"><a href = "http://evg.ulyssis.be/evg/"><b>Eline Van Geert</b></a></br><a href="mailto:eline.vangeert@kuleuven.be" style="display:inline-block; float:middle;"><span style="font-size:5;">eline.vangeert@kuleuven.be</span></a></div></br>

<div style = "float:middle; margin-top: -20px; ">
<a style = "font-size: 16px;" href="http://gestaltrevision.be/en/" target="_blank" rel="noopener">
<span>GestaltReVision research group</br>Laboratory of Experimental Psychology - KU Leuven</span>
</a></br>

<div>
<center><b>Thanks to</b></center>

<center>my research group, institution, and funder</center></br>

</div>

???
I want to thank you all for your attention. If there are any questions, I am more than happy to answer them :-)
If you want to know more, feel free to talk to me during the conference or to send me an email.