Harnad, S. (2003b) Categorical Perception. Encyclopedia of Cognitive Science. Nature Publishing Group. Macmillan.
Differences can be perceived as gradual and quantitative, as with different shades of gray, or they can be perceived as more abrupt and qualitative, as with different colors. The first is called continuous perception and the second categorical perception. Categorical perception (CP) can be inborn or can be induced by learning. Formerly thought to be peculiar to speech and color perception, CP turns out to be far more general, and may be related to how the neural networks in our brains detect the features that allow us to sort the things in the world into their proper categories, "warping" perceived similarities and differences so as to compress some things into the same category and separate others into different categories.
Pullum, Geoffrey K. (1991). The Great Eskimo Vocabulary Hoax and other Irreverent Essays on the Study of Language. University of Chicago Press.
Though I agree that categories could be created this way, this leads back to the problem of elementary category grounding. How many elementary categories does someone have to ground to be able to create all possible (or even most possible) higher order categories? I would assume that the number is rather large. This leads to the question of how someone at a relatively young age could ground enough elementary categories to acquire a sufficiently large amount of elements to recognize everything around them.
ReplyDeleteHow many categories do you need to have grounded directly in order to be able to acquire all the rest through by recombining them into subject/predicate propositions? About 1000:
DeleteVincent-Lamarre, Philippe., Blondin Massé, Alexandre, Lopes, Marcus, Lord, Mèlanie, Marcotte, Odile, & Harnad, Stevan (2016). The Latent Structure of Dictionaries TopiCS in Cognitive Science 8(3) 625–659
I don't think the second question you asked is even relevant. By "relatively young age" I presume you mean children, but do we ever assume children recognize everything around them? They can recognize objects, sure, through learning functions (i.e, a chair, you sit in a chair) and such categories, like animals, but do we presume children to understand what "democracy" means?
DeleteI believe the point of elementary category grounding is to understand that we can can have a basic set of grounded symbols and use them as building blocks to create an infinite amount of categories. In class, Prof. Harnad stated that these sets of around 1,000 categories can be unique to individuals (in space and time). It goes back to the circularity of dictionaries, which by the same logic, is how we create categories. It's dependent on what your set of elementary grounded categories are.
This building block idea is also furthered in the reading through discussion of Boolean combinations. I think children have a surprisingly complex ability to categorize at a very young age, as our brains seem to be wired to efficiently categorize things. This skill would be quite necessary for children, evolutionarily. They may not have all 100 necessary to combine into complex ideas such as democracy, but I don't believe they are so inept at recognition as you're implying.
DeleteThe suggestion is not that children actually ground 1000 categories directly and then learn all the rest by hearsay. We go on grounding categories directly all our lives, even though it could theoretically be replaced by recombinatory verbal descriptions. But the nuclear power of language lies in the fact that that is possible in principle, even if it is not done in practise.
Delete"The nets accomplish this by selectively detecting )after much trial and error, guided by error-correcting feedback) the invariant features that are shared by the members of the same category that reliably distinguish them from members of different categories; the nets learn to ignore all other variations as irrelevant to the categorization"
ReplyDeleteI found thinking about the network bias of computational models to be particularly illustrative of the way that categorization may be accomplished by our brains. While little is known about the brain mechanisms behind categorization, given what we know about neural encoding of memories, it seems like these models provide a window into recognition.
Having a theoretical model for categorical perception is the first step to designing hypothesis driven studies, in contrast to the localization studies critiqued by Fodor. These considerations reminded me of the FFA studies of Greebles, where subtle categorization can be learned. It seems that testing these learned and arbitrary categories will need to be the first step before we can start questioning how innate categories may function.
What do we know about how neurons encode memories?
DeleteOn the other hand, "deep learning" ("neural" net) models do tell us at least one way that a lot of powerful learning could be done.
"In the case of innate CP, our categorically biased sensory detectors pick out their prepared color and speech-sound categories far more readily and reliably than if our perception had been continuous. Could something similar be the case for our repertoire of learned categories too?" (Harnad, 2003).
ReplyDeleteI think it is possible to more readily pick out certain learned categories that we are more exposed to. For example, I would more readily recognize and categorize something as a dog than I would be able to categorize some discrete insect. I would argue that with more exposure to a stimuli, the recognition of that stimuli as part of a certain category is similar to innate categorization.
On a separate but related note, I was thinking about this website called Akinator that I used to go on as a kid. It's essentially a 20 questions game with the computer; that is, the computer asks questions and based on how you answer, it narrows down its options and eventually guesses the correct answer. I wonder if this can be considered as a representation of categorization in a 'toy' model? The computer needs to take your input and uses it to categorize what it is that you are thinking of based on, I suppose, pre-loaded categories given to it.
"Unsupervised" learning by repeated exposure works when the categories are already fairly distinct, but if they are hard to distingusih (or could be categorized in many different ways), "supervised" reinforcement learning by trial and error, with corrective feedback, is necessary.
DeleteTwenty questions may work in ways similar to the dictionary, if most of our categories are encoded verbally.
Re: Motor theory of speech perception
ReplyDelete“According to the (now abandoned) motor theory, the reason we perceive an abrupt change between ba and pa is that the way we hear speech sounds is influenced by the way we produce them when we speak” (Harnad, 2003).
Motor theory was used to explain how and why different speech-sounds are perceived categorically. For instance, according to motor theory we hear either /ba/ or /da/, and nothing in between, as we are anatomically incapable of producing anything between as /ba/ is a bilabial speech sound, whereas /da/ is alveolar.
I am curious as to why motor theory is now abandoned. Harnad refutes motor theory’s assumption that CP is a result of learning to produce speech, as both infants and chinchillas are born with speech CP, and while this may be argued it is to prepare infants to speak as they mature, chinchillas never learn to speak.
However, I am wondering as to whether there is still an underlying motor contribution to categorization that is being overlooked. Galantucci, Fowler, and Turvey (2006) review the motor theory of speech perception and find support for the motor theory’s claim ‘the motor system is recruited for perceiving speech’. For instance, they highlight a study by Kerzel and Bekkering (2000) in which participants were shown a video of a face mouthing /ba/ or /da/. Following the video, they were shown one of two symbols that they had been taught to associate with /ba/ or /da/ respectively, and had to respond to the symbols by producing the corresponding speech-sound. Despite that they were told not to pay attention to the video as it was irrelevant, their response latency was increased on incongruent trials (i.e. when they had to respond /ba/ and the video mouthed /da/ or vice versa), and decreased for congruent trials. This suggests that there is a production-perception link present for speech. Perhaps motor theory does have some merit to it and thus should not be abandoned altogether but rather revised?
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2746041/
The Fowler/Turvey experiment (and theory) is a bit far-fetched, but I too ("Stevan Says") think that the motor theory is down, but not out.
DeleteTo categorize is to do the right thing with the right kind of thing. "Doing" is motor. Assigning arbitrary names is "digital," but other things we do with things are analog (based on the shapes of the objects and our own shapes, manipulating them). Fowler/Turvey make a bit of a cult of GIbsonian "affordances," but there's some truth there too. And don't forget about imitation, motor neurons, and the gestural origin of language...
I too see some value in the motor theory and do not believe it should be totally abandoned. Some fmri studies show how speech muscles get activated when we hear speech.
Deletehttp://onlinelibrary.wiley.com/doi/10.1046/j.0953-816x.2001.01874.x/full
This is interesting because it means that we not only use motor patterns when we are talking ourselves, but that we also rely on these internal motor cues when we categorize others’ speech.
“To show that it [the Whorf Hypothesis] is a full-blown language effect, and not merely a vocabulary effect, it will have to be shown that our perception of the world can also be warped, not just by how things are named but by what we are told about them.”
ReplyDeleteIn youtube video embedded on this page there is an example of individual’s perception of the world being warped by what is told to them. Lera Boroditsky makes the distinction between languages that use cardinal directions to talk about the world, and languages that use body centered directions to talk about the world.
In languages using cardinal directions, the speakers must always be oriented to speak properly. Due to this they develop incredible capabilities of orientation. On the other hand, in body centered direction languages speakers aren’t that competent at being oriented. Thus, depending on how the speaker’s language tells them about the world, they have a different perception of it.
Have a look at egocentric vs. allocentric spatial ability (and stages in children).
DeleteIn a similar vein–although it's not about perception, so much as it is about recall–work by Loftus and Palmer in the 1970's showed that leading language could be used to alter perceptual elements of recalled events.
DeleteObviously that's not as powerful as shifting perception of the world as we experience in real time, but it's a step in that direction.
I wonder, do they really have a different perception of the world? Or is the way they talk about it different?
DeleteI think the real thing that alters our perception of the world is the experiences we've had not the language we use to talk about these experiences. Language may alter our behavior i.e. being able to orient our bodies in certain directions, but I don't think it inherently defines our experiences. What you call north I may call the towards the mountains (or something) but we are both talking about the same thing in the world.
"Berlin & Kay (1969) showed that this was not so: Not only do most cultures and languages subdivide and name the color spectrum the same way, but even for those who don't, the regions of compression and separation are the same. "
ReplyDeleteThis section of the reading reminded me of studies in which researchers found that populations segregated from other humans/technology etc categorized and recognized facial expressions in the same way as other populations globally. This would suggest that reading facial expressions is an innate categorization in a similar way to colour categorization. So perhaps there are some categories that are learned and others that are innate, such as colours and faces. Perhaps it does not have to be an all or none phenomena.
We could also say that all categories are actually learned. They are only innate in the sense that they have already been learned in our evolutionary past. During that "era", those categories we now take for granted as being innate underwent the same process of trial-and-error. Faulty categorizations of facial expressions or of colours could've had grave consequences for organisms or something. And if all people can make above chance categorizations of facial expressions, I'm sure there's still much more room to improve. Anyway, the difference is that the knowledge for those categories is stored in our genes, while all the other categorizations we make now (e.g. a democratic vs undemocratic country) aren't.
DeleteYes, there are both innate and learned categories (but most are learned).
DeleteYes, evolution is a bit like learning (especially Baldwinian evolution), but not quite..
I'm curious to discuss Liberman's idea of "categorical perception." I think it's a valuable observation, but I'm wondering if it isn't just limited to speech. There is a similar (same?) phenomenon in music as well, where we as humans can discern temporal and pitch differences at specific minute increments (otherwise, they sound continuous). Is this categorical perception at work, or a similar but inherently different mechanism?
ReplyDeleteYes, there is CP for pitch (semitones), both relative and absolute.
Delete"CP seems to be a means to an end: Inputs that differ among themselves are "compressed" onto similar internal representations if they must all generate the same output; and they become more separate if they must generate different outputs."
ReplyDeleteThis is an interesting and effective way to explain this. CP as a means to an end highlights the fact that our perception is a tool for us to better understand the world around us. One may think of it as a fixed (human) attribute, where our categories are fixed and-whether innate or not-they simply grow as we develop. This view however entails that they aren't fixed, that they malleably change according to context. In the end, the brain is simply attempting to make sense of the things it perceives, and to do so it must arrange and rearrange its existing 'mental structures' in ways that permit it to accentuate or fade out relevant/irrelevant characteristics, rather than having a different category for every possible experience imaginable (which would be far from efficient).
Learned CP is malleable, but innate CP is not.
DeleteProfessor I was under the assumption that all CP is malleable, or what I'm actually referring to is adaptive (could these words be synonymous? or is there something specific about the term malleability), aka it does not produce the same output across time, every time; in addition that categorization is dynamical thus why would innate CP not be malleable? If for example, with the introduction of smartphones there reaches a time where face to face interactions are kept to an absolute minimum or for the sake of argument we no longer have face to face interactions, we deal entirely in text messaging and social media platforms, over time the advantage of being able to recognize facial expressions (universal emotions) is lost and thus there would be a generation of children with much weaker facial expression recognition than our current generation, and thus an innate CP would become malleable; I would appreciate a counter argument or information on why my thought experiment does not work
Delete"although the primary color and speech-sound categories are inborn, their boundaries can be modified or lost as a result of learning....secondary boundaries can be generated" - I thought this portion of the text was stating that innate categories could be modified? Or at least the boundaries of them can be?
DeleteIn response to Samer: I am not sure your example works in this case, is it really the CP which is malleable when we are talking about change over generations? Just that that specific CP dies out over generations of it being less adaptive - but the idea that an individual's innate category could shift, as suggested in the literature, is a different case.
DeleteRe: Categorical Perception
ReplyDeleteIn the texture dichotomy experiments outlined by Harnad (2003), participants must learn to distinguish between two easy categories of textures (where it becomes obvious after just a few trials) and two hard ones (where only half of the subjects manage to learn the dichotomy). When analyzing the subjects’ performance, we see categorical perception occurs for the hard textures. That is, the two hard categories become compressed within each respective category, and expanded between them. Thus, the learning of these hard categories actually made the categories then look different. For the easy stimuli however, this categorical perception didn’t occur.
I am then wondering why categorical perception (which is not simply categorization but rather a phenomenon that alters the way in which we view the categories through means of expansion/compression) only occurred for the more difficult stimuli and simple categorization was sufficient for the easy stimuli. Is there a categorization-difficulty threshold necessary for us to employ categorical perception? If we were to generate additional conditions for the experiment with ‘medium’ difficulty levels that lie between the easy and hard trials, at what point would we see participants engage in categorical perception?
My guess is that CP separation/compression occurs when you really need it, to reduce the uncertainty: when the categories are highly inter-confusable. You don't need CP to distinguish mountains from valleys...
DeleteLearned CP is probably a side-effect of learning to detect distinguishing features ("invariants") and ignore non-distinguishing features.
This article explores the question of whether or not categorical perception (CP) is an evolved or learned behavior. Although I can see how both positions have their merit, I think that a learned perception of CP most accurately accounts for much more varied types of phenomena while the evolved perspective can be more useful in describing the underlying process behind the learning. We are definitely born with the capacity to categorize aspects of the world. As the articles for this week demonstrated, using categorical perception is an efficient and highly adaptive way for us to organize the world around us, resolving the “blooming, buzzing confusion” which would come if it did not happen. We do not know however, what is going to be important for us to put into categories and that is where the evolved perspective no longer is useful and the learned perspective is. Some people develop categorization skills that others simply do not have based on necessity or interest.
ReplyDeleteOne question that I do have about learned CP is can it account for varied categorizations of the same thing? For example how does the model account for people learning multiple languages that have different phonemic boundaries between sounds? For instance, in English we have a phonemic boundary between l and r but in Japanese the boundary is not the same as it is in English such that these second language learners of English often have trouble with discriminating these sounds. Eventually though after some time speakers can do this a lot better and still speak their native language as they did before. Evidently categorization had to change here but how could it have stayed the same at the same time?
Evolution is lazy: It does not like to hard-code what can predictably be learned.
DeleteThe same (infinite) set of things might be categorized in more than one way, depending on context (the context of inter-confusable alternatives) and what feedback indicates is the right and wrong thing to do in that context. The context could be signalled by anything from a discriminative stimulus (being chased by a predator, don't eat edibles, rather stomp on them!) or, in the human case, words: "Now I want you to sort these cards by color; now by number; now by suit..."
So you don't have to go to multiple languages for examples like this. But, yes, multilinguilism is another example: in the French context you call it une genouille, in the english one, a frog. Ditto for how to pronounce "car."
Higher order categories (fruits, vegetables) sort differently from lower order ones (apples, oranges): Same things, but the right thing to do with them depends on context.
There are an infinite variety of things in the world, and there's an infinity of different ways to sort those same objects.
According to the Whorf Hypothesis " colors are perceived categorically only because they happen to be named categorically: Our subdivisions of the spectrum are arbitrary, learned, and vary across cultures and languages"... I understand that there are opposing views to this such as that proposed by Berlin & Kay "Not only do most cultures and languages subdivide and name the color spectrum the same way, but even for those who don't, the regions of compression and separation are the same." However, how could this be used to explain color blindness or blindness in general? Those who are color blind are unable to percieve differences between certain colors. Do they still categorize them the same way we do? Do they see this fuzzy boundary that Blake and Kay describes or do they follow Whorfs Hypothesis where they categorize according to what they perceive, in which they wouldnt percieve boundaries between the different colors. Those who are blind can perceive and categorize through tactile and other sensory means but color blindness is interesting and I wonder if there are different mechanisms in place.
ReplyDeleteNo, color-blind people have no color CP and blind people can't see colors at all.
DeleteHow can we know that a category is innate or learned? For example, how do we know that the human brain is wired to seek colours in the same way that it is wired to seek human faces? If we were never exposed to colours, I do not believe we would experience any developmental deficits in the same way that a person never exposed to human faces would. There are people with varying levels of colour blindness, even complete black and white vision (although this is very rare) that should by the nature of being humans have innate colour categories, but cannot identify colours in the way most people do. I think it is a bold assumption to believe that colour categorization is innate. Invariance across cultures is a good staring point for distinguishing innate categories from learned categories, but there are always exceptions. Perhaps innate/learned categories could be evaluated at a neurological level to answer this question.
ReplyDeleteColor-blind people do not see colors as people with normal color vision do. Properties can be innate (i.e., coded by genes) but there could be anomalies in which something is different or wrong.
DeleteThis comment has been removed by the author.
ReplyDelete“CP occurs whenever perceived within-category differences are compressed and/or between-category differences are separated, relative to some baseline of comparison. The baseline might be the actual size of the physical differences involved, or, in the case of learned CP, it might be the perceived similarity or discriminability within and between categories before the categories were learned, compared to after.”
ReplyDeleteHow, why and when do these compressions or separations occur? If it is “perceived”, does this mean compression and separation is subjective to each individual’s perception? If so, the argument for these being innate is weakened I think, because innate mechanisms tend to be more reflexive than contemplative e.g. a mother reacting to her baby crying: she does not really think about whether the baby is crying, she just hears it and reacts to move towards it. Yet there seems to be a contemplative element to deciding categories as we compare and consider it’s alikeness to something, maybe for some things this contemplation is super quick like knowing a laptop and bird are different but it’s not reflexive to determine this differentiation. It still incorporates what we learnt through exposure to these different objects, to know that they are different. I don’t think we were born with that capability, rather through visual, auditory and utility comparison of things, we learn to differentiate. We weren’t born knowing what they look like, sound like or do – we learn and THEN compare – again all of which involves contemplation and not reflexes. Maybe through increased exposure our contemplation and categorisation becomes faster and it can seem like a natural phenomenon to differentiate.
“The generic human face may be an innate category for us, perhaps even the various basic emotions it can express, but surely all the specific people we know and can name are not. “
ReplyDeleteI think the idea that some category detectors seem to be innate whereas others are shown to be less so is an interesting concept to explore. First off it shows that functions and categorizations have differing levels of importance. For example as this passage points out, detection of emotions is likely something that most people use in their daily lives, and its importance can be further demonstrated by the problems caused by its absence. If someone is not able to identify the emotions that other people are conveying they are not as capable of functioning smoothly in society. In certain situations these individuals may misread and misinterpret the situation thereby making it worse. As well we have certain abilities to identify faces and identify the faces of people we know. Where this branches off from innate ability (and thereby showing that it is not as important evolutionarily) is the skill of identifying a certain face- for example my ability to recognize someone else’s moms face is not as important as knowing how to tell that I’ve seen that face before).
It's my understanding that our means or "affordances" of categorization are inextricably tied to our sensory-motor systems, which are constant throughout the human species. This is where categorization begins, from animal categorization (before language) to language-induced CP. This evolved CP is the basis for language-induced CP, this is evident when we compare the linguistic categories shared by languages across the world. As was noted in this week’s other lecture “to Cognize is to Categorize”, Chomsky’s Universal Grammar rule of linguistics reflects human means of cognition and expression. Our means of expression is itself tied to our perceptual systems, just as our color boundaries are pre-defined by our retinal ganglion cells, blue opposed to yellow, red opposed to green. As for language-induced CP, this is a result of making sense of stimulus that is produced differentially by many people (correction for poor articulation). We have learned through experience to inadvertently differentiate between voiced and unvoiced consonants and create this sharp boundary between pa/ba. Some of us may do this by paying attention to different aspects of the auditory scene yet we achieve the same result. It can be said that the rules we use to differentiate between phonemic categories may differ, and this stands in contrast with our visual system, but end goal of between category expansion and within-category compression is achieved. This where the Whorfian arbitrary subdivision comes in, but isn’t that how all language is created?
ReplyDelete"Some of these categories are "prepared" in advance by evolution"
ReplyDeleteThe ideas of categories being prepared in advance by evolution is one that is extremely interesting. I wonder, as the world progresses and as our understanding of how genes change over time in a short term (ex: epigenetics) and long term fashion becomes more complicated, if we will understand how new "categories" might be delineated neurally. For example, if there is a new food demarcation with pollutant X (toxin), gives a small smell to food it's in, would it create a new category for (deadly toxin smell or not deadly toxin smell). How important does a category have to be, and what qualities would it need for it's demarcation to become "innate" and not require any special learning?
I have a slight issue with the point brought up under the heading of "The motor theory of speech perception". Lieberman et al 1957 did report that people could only hear /ba/ and /pa/ and nothing in between, but these themselves are distinct categories are learned based on which language you speak. From birth, infants are exposed to a language which will ultimately determine what types of phonetic contrasts they are sensitive to. At about 12 months of age, babies are only able to pick out the phonetic contrasts that are relevant to their language but before this time they are sensitive to phonetic contrasts of other languages. For example, French monolingual 1-year-olds are able to distinguish between /ba/ and /pa/, or /ba/ and /pa/ where the p is aspirated, but not between an aspirated p and an un-aspirated p. English monolingual babies were able to distinguish between /ba/ and /pa/ with an aspirated p, but were unable to distinguish between /ba/ and /pa/ (un-aspirated). Bilingual French-English babies, however, were sensitive to all phonetic contrasts. Would this not be evidence for the Sapir-Whorf hypothesis? The categorical perception of an individual is actually based on what language they speak.
ReplyDeleteThe Whorf Hypothesis states that the way we interpret colours is dependant on cultural language differences. The perception of colours is categorical because they can be named categorically according to the arbitrary spectrum subdivisions. However, I disagree with this hypothesis given that just because one culture names a category differently doesn’t mean that it is an entirely different category altogether. Additionally, human eye anatomy and visual system organization in the brain (save for morphological abnormalities and pathologies) is more or less the same; we have the same amount of rods and cones and the same structural organization, therefore there is no reason that we would interpret colours that differently. I agree with Berlin and Kay’s debunking of the Whorf Hypothesis.
ReplyDelete"The motor theory of speech perception explained how speech was special and why speech-sounds are perceived categorically: sensory perception is mediated by motor production. Wherever production is categorical, perception will be categorical; where production is continuous, perception will be continuous."
ReplyDeleteAccording to the motor theory, the way our mouth/nose anatomy limits what we can produce, and therefore what we can perceive. Although the motor theory has now been abandoned, I think we can still deduce that how our body interacts with the world in a sensorimotor way is connected to our ability to categorize. I wonder whether this means that categories are separate because of the restrictions of our anatomy, or whether over time human anatomy has evolved to adapt to separate categories. I guess this is sort of a chicken and the egg situation: what came first?
“Some of these categories are "prepared" in advance by evolution: The frog's brain is born already able to detect "flies"; it needs only normal exposure rather than any special learning in order to recognize and catch them.”
ReplyDeleteIf frogs were hypothetically raised by snakes who only eat crickets, could they move “flies” from the category of prey to the category of “unimportant other being” despite being already “prepared” in advance by evolution? If we were to expose flies to these frogs after they were already habituated to eat only crickets, would the brain’s innate “prepared” categorization still be able to come out and give the flies the ability to recognize and catch them with simply exposure, or would special learning have to be in place to give them these abilities? In other words, would it be possible for the innate categories we are born with to be altered if raised in the right environment, or are they hard-wired into the brain no matter what?
In response to your comment I thought of this section in the article: "although the primary color and speech categories are probably inborn, their boundaries can be modified or even lost as a result of learning, and weaker secondary boundaries can be generated by learning alone"
DeleteSo, to answer your question "would it be possible for the innate categories we are born with to be altered" - yes, to some extent it would seem there is room for some alteration. That being said, Stevan says somewhere above in the comments that innate CP cannot be modified…. So maybe I've missed something!
“It is not that all pa's are indistinguishable and all ba's are indistinguishable: We can hear the differences, just as we can see the differences between different shades of red. It is just that the within-category differences (pa1/pa2 or red1/red2) sound/look much smaller than the between-category differences (pa2/ba1 or red2/yellow1), even when the size of the underlying physical differences (voicing, wave-length) are actually the same.”
ReplyDeleteI am wondering if there’s a natural tendency that the ability of detecting between-category differences is usually innate or can be learned by mere exposure, and within-category differences are generally more difficult to detect and require supervised learning?
I think the concept of CP for colors to be fascinating. As a counter point to the Whorf hypothesis Dr. Harnad points out that certain colors, RGB, are innate. However, he concedes that other colors, scarlet and crimson, and fuzzy concepts like them, are likely learned. I was wondering what aspects of culture may cause certain cultures/languages to have more categories within a range in the color continuum. For example, more nameable colors between blue and purple. Is this evolutionarily driven? Or, perhaps certain cultures had more abundance of certain pigments within their holdings. E.g. purple dye is associated with royalty because of the difficulty of harvesting the pigment from snails in the Mediterranean during the Classical era. Further, perhaps conceptual words, the "goodness, truth, and beauty" words of our vocabulary, have similar cultural origins. My Korean friend has often expressed to me that Korean there are far more words to express emotion than in English. Also, perhaps regions/cultures that have been war torn for centuries (Israel/Palestine) have more words for violence/warfare? Does this actually matter though? Dr. Harnad points out that we can simply use more words to express these categories that are contained within one word. Why do certain cultures then feel the need to assign a word/category to a concept, and others simply summarize it in numerous words?
ReplyDelete“[…] to show that [the Whorf Hypothesis] is a full-blown language effect, and not merely a vocabulary effect, it will have to be shown that our perception of the world can also be warped, not just by how things are named but by what we are told about them.”
ReplyDeleteThe Whorf Hypothesis (weak version) states that language has an effect on the way we see the world. I think that there are various examples showing that what we are told about things affects our perception of the world. An example from the medical field is the way in which we conceive of the same disease across cultures. In Western countries, mental illnesses such as depression are increasingly being recognized and treated as health issues on the same level as other physical illnesses. However, in some parts of the world such as Asia, it is more taboo to seek professional help for issues related to mental health, and therefore there are significantly less diagnoses of depression. Instead, people who suffer from depression are often told that the way they feel is an indicator of their weakness in dealing with life’s stressors. In short, because people who suffer from mental illnesses are told different things about who they are and what they are capable of depending on the culture they are embedded in, they perceive themselves and the world differently.
I wonder if there’s a discrepancy between how you’re using the word “perception” and how Harnad is using it. Our understanding of concepts might vary based on our beliefs (a contemporary example would be millennials who’ve problematized the tv show Friends based on changing expectations and acceptance of jokes using gender and body weight tropes for example). On the other hand, if we take perception to be the processing of sensory inputs, the question becomes “can language literally change the way the world looks to you?”
Delete
ReplyDelete“That is enough to rehabilitate the Whorf Hypothesis from its apparent failure on color terms (and perhaps also from its apparent failure on eskimo snow terms, Pullum 1989), but to show that it is a full-blown language effect, and not merely a vocabulary effect, it will have to be shown that our perception of the world can also be warped, not just by how things are named but by what we are told about them.”
Studies on bilinguals have shown that it is very possible that what we are told about things or the way we address them changes our perception of them. For example, a study has shown that because of different gendered determiners in front of the word ‘key’ speakers of different languages associated different things with the object. A feminine determiner allowed the key to associated with more feminine traits such as small and delicate, whereas a male determiner allowed it to be related to strength and sturdiness. While the Whorf hypothesis is mistaken with how words shape our perception of the world, it seems that the language we learned a specific term in could still affect our perception of what we think of about said term.
Liberman’s take on the Motor Theory of Speech Perception can be alternatively stated as a discordance between a continuous element and categorical perception. The theory states that auditory perception of speech is influenced by the method by which we produce speech. The discordance arises from the fact that the voice-on-set time is continuous as it varies along a continuum from “ba” to “pa”, yet these speech sounds are perceived categorically as either being “ba” or “pa”, depending on its proximity to either along the continuum. Nonetheless, this theory has been falsified and wrongfully describes this perception as categorical since the claim is essentially based on “likeness” or relative judgement. That is, the Motor Theory of Speech Production essentially attributes the perception of “ba” or “pa” in the context of this varying continuum of voice-on-set-time as being dependent on the comparison of the articulated speech sounds proximity to either “ba” or “pa.” In Cognition is Categorization, Harnad describes categorization as absolute discrimination, which is far narrower than the relative discrimination proposed in Liberman’s theory, since it requires identifying the element in isolation.
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteThis paper does not seem to come to a firm conclusion on weather speech sounds are innate or learned. Liberman found that when people listen to sounds along a continuum they hear “ba” and “pa” alone without the sounds in between. This would support the innate hypothesis and explain categorical perception (CP), when the jump is from category to category rather than changing gradually. So, since “ba” and “pa” are distinct it means that they are categorical perception. But, according to the “motor theory, the reason we perceive an abrupt change between ba and pa is that the way we hear speech sounds is influenced by the way we produce them when we speak”, suggesting these categories are learned through experience of learning language.
ReplyDelete“full-blown language effect, and not merely a vocabulary effect, it will have to be shown that our perception of the world can also be warped, not just by how things are named but by what we are told about them. “
ReplyDeleteThis ending to the paper made me think about the Whorfian hypothesis from a new perspective. I found that the examples of the man and the bachelor were perfect in describing how different terms can compound to form higher order combinations. Although I had learned about the Whorfian hypothesis in many other classes, I had never fully considered it from this point of view. After having read this section of the article, I must say that I can now understand and am convinced that using certain terms could help one have less or more information, and change the way about which they talk about something. For example, it is very clear that ‘bachelor’ gives more information about an individual than ‘man’, and that if I was told about someone who was a ‘bachelor’, I would think about them very differently than if the same person had been called a ‘man’. In this way, I can see how the way we think is at least somewhat influenced by the language we use.
“Humans have such innate category-detectors too: The human face itself is probably an example.”
ReplyDeleteWhen reading this section about innate category-detectors, it reminded me of something I had read regarding people who have autism. It said that often time people who have autism are largely impaired in face processing, memory for facial identity, and recognition of facial expression of emotion. The study reported that the subjects they looked at had abnormal activation in certain brain structures, which is what they assume led to the lack of facial recognition. I was wondering then, if certain categories are not necessarily innate in humans, but instead innate based on the way that we consider an “impairment-free” brain to function/fire? In addition, I was thinking that the human face probably is considered a category for most people because it is something that we prioritize as important because it aids in social interactions. Could it be possible that the level of relevance we give a certain feature can turn it into a category?
Also, if you want to see the study: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2174902/
There are even recent demonstrations that although the primary color and speech categories are probably inborn, their boundaries can be modified or even lost as a result of learning, and weaker secondary boundaries can be generated by learning alone (Roberson et al. 2000).
ReplyDeleteThis part of the paper addresses the Weak Whorfian Hypothesis that we discussed in class which was: the way that we learn to categorize things (which may or may not be a result of language) can affect the way we perceive the world. An example of this would be the study that was done between Russian and English speakers and their perception of the different shades of blue. Russian makes a distinction between two types of blue (light blue and dark blue) while English only has one word to describe blue. They were put through a task to match squares of two different shades of blue to a target square. The results found a significant difference in reaction time between the English and Russian speakers, which shows that even though colour categories are innate, their boundaries are able to be modified.
It was interesting to read about this as I was a participant for an EEG study conducted in Prof. Harnad’s lab that was investigating the impact of corrective feedback on learning new categories. I did a task in which I had to listen to three sounds and choose if the third sound was the same as the first or the second sound. I then had to classify different sounds into category 1 or category 2 and I received feedback about whether I had sorted the sound in the right category or not.
ReplyDeleteIt was interesting to see a brief summary of my results as I had learned how to correctly categorize the sounds. After the debriefing, I learned that there is an EEG signal that is associated with having learned a category. It makes me wonder why some people are able to learn these categories and some people aren’t. Also, is there a critical period for learning categories? Or is all that matters the amount of corrective feedback you get when you are in the process of learning that category?
"Computational modelling has shown that many types of category-learning mechanisms display CP-like effects."
ReplyDeleteDoes this statement mean that we can only create machines that that perceive categorically, or can a machine also perceive continuously. Although I don’t know the answer, I imagine that a machine could compute continuously on a macro level, but based on its binary language of ones and zeros, it would be categorizing discretely on a micro level. Is the problem of computationalism really just a problem of figuring out how to achieve continuous computation in place of discrete, or are these two binaries completely unrelated?
"Can categories be acquired through language alone?"
My impulse on this question is yes. I feel like the very process of learning a language is learning the categories and how to classify individuals to them. In this way, I think much of what we know of categories we know through language. Though I agree that there are certain categories that are with us since birth, there are others that I would postulate are developed entirely through our language. For example, higher-level categories such as the distinctions between lakes, rivers and oceans or how to discriminate a safe plant to eat from a poisonous one seem to me to be entirely rooted in verbal description. I wonder how you would set up an experiment to test this verbally quality of categories.
"Does this statement mean that we can only create machines that that perceive categorically, or can a machine also perceive continuously."
DeleteI think that the CP-like effects being talked about are observed only specifically in machines whose purpose is to learn categories. I think the point is that CP is a feature of categorization. If we perceived everything as continuous, we wouldn't ever be able to make the distinctions required to put something in or not-in a category. The act of grouping necessarily means that we're perceiving some things are more similar than others. There are machines which "perceive" continuously (although I don't know if perceiving is really the right term). Take the example of pitch detector. If it's not deciding whether the tone is more like C or C#, it can give you the pitch in Hz, and it can give you a number for any pitch on the spectrum.
"Although I don’t know the answer, I imagine that a machine could compute continuously on a macro level, but based on its binary language of ones and zeros, it would be categorizing discretely on a micro level. Is the problem of computationalism really just a problem of figuring out how to achieve continuous computation in place of discrete, or are these two binaries completely unrelated?"
This is a very interesting point. All information is fundamentally binary. Things either are or they aren't. However, I don't think this is a flaw of computationalism. The problems raised in the CRA notwithstanding, we don't perceive stimulus intensities as continuous, so why should a computer have to? And most computers are powerful enough to approximate continuity to an inscrutable degree.
"Can categories be acquired through language alone?"
The answer is yes, provided you have enough categories to begin with. I don't think the categories need to be innate beyond some very basic ones (consider Prof Harnard's example of mountains and valleys). If you can tell very basic differences between certain things, you can acquire the minimum needed categories through sensorimotor interaction with the world (i.e. trial and error). If you've acquired enough of these categories, they can be recombined to acquire all others.
““warping” perceived similarities and differences so as to compress some things into the same category and separate others into different categories”
ReplyDeleteIs uncertainty reduction a possible function of CP? Since the paper notes “the nets learn to ignore all other variation as irrelevant to the categorization”. Does the answer to that depend on whether it is learned or evolved CP? It could be possible that we employ CP because we don’t like being unsure about which group something belongs to and due to our compulsion to simplify as a means of reducing cognitive strain when thinking about something since knowing something is of a “kind” gives us more info about it than knowing it as just an individual. I think uncertainty reduction could also be connected to, as the article mentioned, the diminished perception of intra-category disparities as opposed to inter-category ones.
It’s also interesting that CP is output based, but what pushes us to create a new category. How different does the output generated by an input have to be to induce creation of a completely new category rather than compressing it into existing ones? How significant does the between category separation have to be and how does the brain determine this? Is the ability to create separation innate? But how does that work for differences we learn because we surely aren’t born knowing all the different kinds of things in the world. Thus far brain imaging has only localised brain regions related to categorisation but that still doesn’t answer these questions pertaining to the how. If areas in the brain light up when having learnt a category successfully and this is absent in those that haven’t learnt it, will innate categories cause the same regions in everyone brain to light up?
This reading begs the question “can categories, and their accompanying CP, be acquired through language alone?” Personally, I believe categories shouldn’t be acquired through language alone, but through the experiences and emotions that we go through. We certainly need language to communicate beyond the categories given to us, such as the language used by animals to express their positive feelings, and to warn other animals about potential threats. After experiencing the same events multiple times, however, the same animals wouldn’t require the use of language to warn the other animals about threats or express positive feelings because they would have all gone through the same experiences and wouldn’t need to inform the others because they all have the knowledge by then.
ReplyDeleteAnother topic I found interesting was the Whorf Hypothesis, in which “colors are perceived categorically only because they happen to be named categorically.” So if we were to not have a name for the colors blue and green, then would we be able to tell the difference between each, since they’re both similar colors? This changes the way we could categorize colors, since our perception of the wavelengths would be different.
"But to show that it is a full-blown language effect, and not merely a vocabulary effect, it will have to be shown that our perception of the world can also be warped, not just by how things are named but by what we are told about them." Well isn't that what happens each time we dream? Dreams are a perfect example of how our perception can be warped depending on our perception of the world. Additionally, I think our capacity for language allowed us to create myths, hence the whole idea of gods and society. These are all abstract categories that language alone allowed us to have. Because of that ability, we are able to create a category for unicorns, which have never existed but somehow someone decided were mythical creatures of great power. Same with religion. It is a mix of language and qualia that enabled us to form categories of imaginary things.
ReplyDeleteThank you for nice information
ReplyDeletePlease visit our web:
Click Here