Inference or perception?
Whether Causation is inferred or perceived is a matter of debate among cognitive psychologists
David Hume said "Causation cannot be measured directly. It is something that we infer from our perceptions." As argued by Hume, we do not have a perception of causation, as perceptions must have sensory causes. Instead we infer causation from our perceptions (such as the low level features described below). The inference of causation is a cognitive belief that one thing causes another.
However Causation could be seen as a Perception, rather than an inference. This is because inference implies a higher process, that is cognitively demanding, which this isn't. Perception of causality is fast, requires little to no thought and automatic, all of which are qualities of perceptions, not inferences.
Factors affecting the inference of causality (Michotte 1963)
There are 5 factors that are said to influences our inference of causation. All examples will be described in terms of A (which 'acts') and B (which 'reacts'). Picture them as shapes in your head, with A on the left and B on the right (if using Tom's lecture slides as a visual aid).
1.Launching: A moves towards B until it makes contact and stops, after which B moves directly away from A. Not in Tom's lecture slides but a key part of the theory.
2.Priority: For A to cause B, A should come first.
3.Timing: once A has acted, B must act within a certain time frame for us to suspect causality. If A moves towards B, and B moves away 4 seconds later we don’t often believe that A caused B to move. Shanks, Pearson and Dickinson (1989) showed that the most important factor in perception of causality is the delay between A and B. However Buehner and May 2004 found that the time delay effect can be abolished entirely if Pps were made aware that random delays were possible.
4.Proportionality: the action of B must be proportional to the action of A if we are to infer that A caused B. If A moves towards B, and B moves away very slowly we may not infer that A caused B to move. If A moves towards B and B moves randomly, or moves and then moves back to where it started we may not perceive causality.
5.Exclusivity: for us to infer A caused B there must not be any other variables that could plausibly have caused B. for example if we introduce C and D it must be very obvious that C and D did not cause B for us to infer causality about A. Schlottman and Shanks (1992) argued that whether A is a good predictor of B or not is important in the judgement of causality; A must reliably predict B.
Note that these factors are sensitive: a small change in delay or speed or reaction or direction of movement can lead to us failing to see causality.
Michottes work is remarkable in that it demonstrates how little info we need to infer causality; simply 2 blobs moving around is all it takes.
White and Milne’s other factors
White and Milne expanded the list of phenomenon that allowed us to perceive causation.
1. Pulling: if A moves and after a tiny delay B moves but thereafter appears to move with A then we will perceive A as pulling B. interestingly this can be offset If A moves downwards, possibly because subjects take gravity into account, complicating things somewhat.
2. Forced Disintegration
Inference of causality as a heuristic
The inference of causality is possibly a heuristic. It can lead to the illusory correlation, or the perception of a non-existent correlation between 2 co-occurring variables.
An example of Causal Inference/Perception
A great example of this is Tim Minchin's instrumental act, where he plays an imaginary guitar, bass and drums in time with background music. Although we can clearly see that the instruments don't exist, Tim's timing of his actions coincide with the music perfectly, so we perceive him to be the cause of the sounds and therefore making the music. This examples how low level perceptual features such as the immediate timing of his actions (timing) and consistency and appropriate force (proportionality) of them in time with the music, result in the inference of a causal relationship between his actions and the music. NOTE While this is a good example it is not evidence and should not be used as such in the exam
This is said to be the reason many superstitions came about, as well as the basis for the halo effect.
Chapman(1967) presented pps with pairs of words and asked them to report which ones occurred together most often (in reality they occurred in all combinations the same amount of times). They found that participants often rated semantically associated words such as ‘bacon’ and ‘eggs’ as co-occuring more frequently than non-related words. That said, it could be argued that this is an example of the availability heuristic, not illusory correlation, as participants rate words as co-occurring because they are more easily paired in the mind; the pairing is more available in the mind.
Chapman and Chapman (1967)investigated the relationship between strength of associative bond between two events, and perceived co-occurrence of the two events. Participants were presented with cases of hypothetical mental patients, given their diagnosis report and the (hypothetical) patient's own drawing of a person. The participants were to judge the frequency with which each diagnosis (e.g. 'suspiciousness','paranoia', etc.) accompanied various features of that patient's drawing (e.g. peculiar eyes). The researchers found that ppts markedly overestimated the frequency of co-occurrence of natural associates (e.g. an actual person's peculiar eyes, and their suspiciousness) within the context of mental patients and their drawings. In other words, the ppts made an illusory correlation between the patient's diagnosis and features of their drawing. Someone who is suspicious doesn't necessarily draw a person with suspicious eyes, but one does associate those two things together.
--> Link to availability heuristic (Tversky & Kahneman, 1974) - the two events are paired easier in the mind, making them more available.
Chapman and Chapman (1969) examined psychiatrists use of the Rorschach ink blot test as a diagnostic tool for homosexuality, and while they found no validity to the test they did find a sample of untrained undergraduates arrived at the same conclusion, even when given random pairings. this can explain why many clinicians can converge on the same, wrong explanation for a phenomenon. Hamilton and Rose 1980: gave Pps descriptions of people involving one of 3 professions (doctor, accountant and salesperson) and also included 2 personality traits. Personality traits were randomly assigned to description from a pool including a mix of stereotypical (for example talkative for salesperson) and neutral traits. Pps reported that stereotypical trait co-occurred with the occupations that they were associated with.
Another example of illusory correlation is where people have lucky possessions: they see a fake correlation between their item and good events happening to them.
Michotte’s results seem to hold up across cultures. A study by Morris and Peng (1994) showed that Chinese and English Pps showed the same inference of causation even though their explanations of how it arose were dramatically different.
A study by Leslie et al showed that 6 month old babies habituated to Michotte’s videos were surprised when the video was reversed, but weren’t when the Michotte’s video did not lead to inference of causality (for example if the video had a large delay between A and B’s actions.)
On the other hand Murphy et al 2011 showed that illusory correlation appears only after exposure to descriptions of people, but then disappears if enough descriptions are given. People start with no illusory correlations, develop them rapidly, and then lose them gradually as they are exposed to more exemplars of the stereotype.
Why do we have these biases?
This over-preception of causality may have evolved simple because its safer to assume that something A caused B and be wrong than to assume A didn’t cause B when it did. It seems to be inbuilt as it develops pretty early: even young babies are surprised when the rules of causality are violated.
Perceptual animacy is where we assign life and life-like qualities to objects and agents. At the most basic level we may perceive something as alive, and at higher levels we may attribute them goals and desires (the tree is trying to grow as tall as possible, and The forest wants us to leave).
Heider and Simmel 1944 showed participants an animation of a series of shapes moving around. Participants reliably perceived the shapes to be purposeful agents, even going as far as to assign them emotions and morally judge their ‘behaviour’ (such as “the big triangle is angry, the little triangle is scared, the big triangle is a nasty bully”) Studies have suggested that animacy arises from the movement patterns of the objects, and less about the spatial relations between them (Berry et al., 1992). Stewart et al proposed an energy violations hypothesis: that if movement appears to violate Newtonian laws then it implies a hidden energy source and probably animacy. Factors that consistently result in the percept of animacy include
1. An object starting from rest
2. Direct movement towards a goal
3. Sharp changes in direction to avoid obstacles
It is argued about how many factors or criteria that need to be met before we attribute animacy. Tremoulet and fieldman somewhat refuted the energy violations hypothesis, showing that simple changes in direction and speed on a featureless background can lead to perception of animacy. Gergely et al have shown that this perceptual animacy is even present in 9 month old babies. Other studies have shown that chimpanzees also are affected.
Using Cognitive Neuroscience to Observe Inference of Causality in the Brain
Causation has been further investigated by Heberlein et al (1998), who carried out a study on a patient suffering with damage of their amygdala in the brain. It was found that this patient did not perceive animacy of objects the same way equivalent control subjects did. Therefore, this acts as evidence for the amygdala to being involved in perception of animacy.
A further neuroscientific study was conducted by Happe and Frith (2000), who used PET scans to demonstrate more activity in the following 4 areas of the brain when subjects were shown an animate display:
- Tempoparietal junction
- Fusiform gyrus
- Occipital gyrus
- Medial frontal cortex
Links to other topics
The Johansson movie is where we see a group of dots moving in a certain way and perceive them as a person moving. This is similar to inference of causality as we infer that these dot movements are actually purposeful, that they are moving together in a coordinated way.
Perceptual animacy is the basis for anthropomorphising.
Perceptual animacy could very well underpin our social instincts; we are programmed to see social interactions.
Is there any evolutionary benefit to distinguishing between true causes and reliable correlation?
Perceiving true causation is the ideal and would certainly provide massive evolutionary advantages. However to gain this perfect ability it would probably require large amounts of time an cognitive effort, so the negatives outweigh the positives. We have evolved heuristics such as illusory correlation to do the job good enough, but most importantly, fast. though it can lead to mistakes, the speed and lack of effort it takes would allow us to outperform an organism that had a 'perfect but slow' strategy.
Scholl, B. J., & Tremoulet, P. (2000). Perceptual causality and animacy. Trends in Cognitive Sciences, 4(8), 299 – 309 http://perception.research.yale.edu//papers/00-Scholl-Tremoulet-TICS.pdf
Anderson, R. B., Doherty, M. E., Berg, N. D., & Friedrich, J. C. (2005). Sample size and the detection of correlation–a signal detection account: comment on Kareev (2000) and Juslin and Olsson (2005). Psychological Review 112(1), 268-79
Fiedler, K. (2000). Beware of samples! A cognitive-ecological sampling approach to judgment biases. Psychological Review, 107(4), 659-676.