The 'Units' of Perceived Animacy
Below are demonstrations of the various conditions and phenomena reported in the following paper:
- What are the underlying units of perceived animacy?: Chasing detection is intrinsically object-based. Psychonomic Bulletin and Review, 24, 1604-1610. PDF .
One of the most fundamental questions that can be asked about any visual process is what sorts of inputs it takes - e.g. features, spatial locations, or objects. This has been studied, for example, in two of the most active areas of research in visual cognition - attention and visual working memory. In both cases, it has turned out that these processes do not operate equally well over any arbitrary input feature, but rather that their preferred input units are visual objects.
We wondered whether perceived animacy
is similarly object-based - might there be a perceptual bias to see only discrete visual objects as alive? Observers
detected chasing under two display conditions: In the Unconnected condition a 'wolf' disc chased a 'sheep' disc
amid several randomly moving distractors. In the Connected condition the wolf and sheep moved identically, but we disrupted the objecthood of both by connecting each to a different distractor via a thin line. These lines did little to obscure the locations or motions of the wolf or sheep, so in principle their chasing movement should have been equally noticeable in both conditions,
Nonetheless, observers were extremely bad at detecting chasing in any condition in which the wolf and sheep were not both discrete objects. This illustrates how
the visual system encodes an "ontological assumption" about what can and cannot be alive - to be alive, you must first be a discrete thing!
This research was conducted in collaboration with Tao Gao.
Experiment 1 — Descriptions
In our initial experiment, we were interested in whether observers would describe chasing events differently when the participants in these events were not discrete objects, but rather object-parts. Observers watched single animations in which a wolf disc and a sheep disc were briefly highlighted, and then chased one another around for a few seconds. In the Unconnected condition, the wolf and sheep were both discrete objects. In the Connected condition, they were each connected to different distractor objects via thin lines. Amazingly, this simple manipulation caused observers to be half as likely to mention mental states when describing the chasing motion. The graph below depicts the rates of mental state language use in the Unconnected and Connected conditions. Click/tap the bars to see sample displays for each condition.Click/tap the bars to view example displays
Experiment 2 — Detection
How automatic is this role of objecthood? Were observers in Experiment 1 really blind to chasing in the Connected condition, or were they just being more conservative with their language? To test whether disrupting objecthood causes us to fail to see chasing altogether, we asked a new sample of observers to try their best to detect both chasing performed by discrete objects and chasing performed by line-ends. Observers were fully familiarized with both types of displays, and they were told that both types of chasing would occur equally often in the experiment. Since this was a detection task, we also needed to include chasing-absent displays. We accomplished this by having the wolf chase the invisible mirror image of the sheep, reflected through the display's vertical midline (mirror chasing preserves low-level motion features, such as the amount of correlated motion between the wolf and the sheep, but it effectively abolishes the chasing percept). We generated mirror-chasing analogs for the Unconnected and Connected chasing displays, and measured observers' ability to discriminate chasing from mirror-chasing in the Unconnected and Connected conditions. The graph below depicts chasing sensitivity (d') in the Unconnected and Connected conditions. (Click/tap the first bar to see an example trial in which chasing is performed by discrete objects, an example of a mirror-chasing trial from this condition, and a "cheat" video that illustrates how the mirror-chasing displays were constructed. Mouse over the second bar to see an example trial in which chasing is performed by line-ends, an example of a mirror-chasing trial from this condition, and a corresponding "cheat" video.) As can be seen from the graph, observers were much more sensitive to chasing performed by discrete objects. This was true despite observers' best efforts to detect chasing in both conditions. In other words, chasing-detection is irresistibly object-based.Click/tap the bars to view example displays
Experiment 3 — Wolf vs Sheep
How powerful is this effect? Is disrupting the objecthood of either the wolf or the sheep sufficient to impair chasing detection? We ran two new versions of our detection experiment. For half of observers, we attached the wolf to a distractor in the Connected condition, and for the other half of observers we attached the sheep to a distractor in the Connected condition. We found that disrupting the objecthood of either the wolf or the sheep was sufficient to impair chasing detection. (Click/tap the bars below to see what chasing looks like when just the wolf is connected, or when just the sheep is connected. As can be seen from the graph, chasing detection was impaired in both of these conditions.) Why must the wolf and sheep be discrete visual objects in order for chasing to be perceived? Perhaps the best explanation appeals to a role of grouping processes in the perception of animate interactions - Seeing chasing requires attending to the wolf and sheep simultaneously (i.e. grouping them). If either the wolf or the sheep is already participating in a different group, then chasing will not be perceived as readily.Click/tap the bars to view example displays