Lecture #10-12 (Fodor Modularity of Mind Part 3)

3 Way Distinction

1. Transducer (register proximal stim)

2. Central Systems (domain general reasoning)

3. Input Systems (interface b/w transducers + central system)

Transducer

Register proximal stim

3 Way Distinction

Central System

Domain-general reasoning

3 Way Distinction

Input Systems

Task of identifying whats in the perceptible environment is “detached” from cog + handed off to input systems; Interface b/w transducers + central system

3 Way Distinction

Perceptual Analysis

Not strictly a species of thought according to this model → the input system is not doing cog BUT is intelligent (just not doing what central system is)

Is input system intelligent?

Yes; Just doing something dif than central system (not cog)

2 Reasons why we Need Input Systems + cant have central processors operate on whats transduced directly

1. Gets things backwards (perception is older than cog, so it must come first)

2. Central processing uses domain-general resources + operates more slowly

Gets things backwards Reason (2 Reasons why we Need Input Systems + cant have central processors operate on whats transduced directly)

Gets things backwards; Perception is older than cog, so it must come first

Domain General Resources Reason (2 Reasons why we Need Input Systems + cant have central processors operate on whats transduced directly)

Central processing uses domain-general resources + operates more slowly → input systems can tell you whats out there without interference from cog

We want perceptual systems to be SEPERATE from central cognition

Ex. If youre thinking about what youre gonna eat for dinner, you dont want that to hamper your ability to see a tiger thats about to eat you

Fodor on Perceptual Accuracy

Perception is accurate enough to navigate the world but not 100%

Ex. Illusion → we can see an illusion + know what we are seeing is wrong but we still cant SEE it that way (this implies modularity)

How do illusions imply modularity?

We can see an illusion + know what we are seeing is wrong but we still cant SEE it that way

When is this model a good way to design a mind?

If perception is generally accurate bc they cant be “corrected” by cog

Fodor’s Key Claim on Input Systems

Input systems constitute a “natural kind” (a category whose members share some underlying, objective nature, even if we can't identify that nature just by looking at them)

Ex. “Water” → a certain kind of liquid, made of H20, drinkable; if we looked at lemonade we dont necessarily know what its chemical compound is

Natural Kind

A real category of things that share important properties because of how they actually are - not just because we decided to group them together + are helpful for generating novel predictions (A scientifically meaningful category whose members share important properties)

A class of phenomena that have many scientific properties in common over + above whatever properties define the class

Ex. There is no single underlying structure for sandwich as there is for water

What does Fodor group with perception as input systems?

Lang (eg. recovering sentence meaning)

Why is Fodor’s grouping of lang with perception as input systems unusual?

Bc lang is typically tied to higher cog (not shared with animals, used in thought)

What part of lang is sort of perception like?

Recovering sentence meaning; You are getting a complex syntactic hierarchy from reading + audition, thus you always grasp meaning even if you try not to → Automatic

Ex. If ants spelled out “I love you” by accident you cant NOT grasp the meaning even if you tried

Fodor’s Claim about Sentences

Just like perception tends to be accurate, sentences tend to be true → thus we can hand it off to input systems

Better Argument for Fodor’s Claim about Sentences

Sentence parsing is a specialized task that calls for a module just like aspects of perception

Modularity

Input systems are modules

9 Characteristics of Modules (some modules may lack some features)

1. Domain Specific*

2. Mandatory (Automatic)*

3. Limited Central Access

4. Fast

5. Informationally Encapsulated*

6. Shallow Outputs

7. Fixed Neural Architecture

8. Characteristic Breakdown Patterns

9. Characteristic Ontogeny (=Development)

Domain Specificity Def

Domain specificity means a mechanism is specialized for processing a particular type of information and/or performing a particular computation

One sensory modality can contain multiple specialized modules (like color perception + 3D shape in vision) + a module doesn't necessarily have to belong to just one sensory modality (for lang can recieve through hearing + vision like McGurk)

Concerns the type of information processed AND the specialized computation performed

9 Characteristics of Modules (some modules may lack some features)

McGurk Effect

A perceptual illusion where visual cues (like lip movements) conflict with auditory cues (like spoken sounds), causing the brain to blend them into a third, unique percept → Shows that a module doesnt have to belong to just one sensory modality (demonstrates cross-modal integration)

Visual input can change what you hear; Even when yk how it works tho it still works

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

2 Properties of Domain Specificity

1. Input Condition

2. Computational Condition

Input Condition

Only a very specific class of representations can be inputs to the module

2 Properties of Domain Specificity

Computational Condition

The computation being performed is specialized to a very specific task

2 Properties of Domain Specificity

Fodor + Domain Specificity

Domain-specificity as an aspect of stimuli; A module is specialized for a particular domain of information → different modules process different aspects of the information we receive

Ex. Color, shape, 3D space, bodily motion, facial recog, sentence parser

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Eccentric Stimuli

A category of stimuli that requires specialized information to interpret

Ex. Color → Your visual system needs to process information about wavelengths and color differences

Ex. Faces → Recognizing faces requires processing specific facial features and their arrangement

Ex. Lang → Understanding sentences requires processing grammatical structure

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Implication of Domain Specificity being a Stimuli

Understanding a particular type of stimulus requires specialized information → Because these tasks require different kinds of information and computations, Fodor suggests they are handled by specialized mental mechanisms, rather than one general-purpose mechanism doing everything BUT ordinary objects (like cow) DONT need this specialization

When a stimulus domain is eccentric—meaning processing it requires specialized information—it's likely to be handled by a specialized computational system

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Do ordinary obj need the specialized mental mechanism?

No

Ex. Cow may be processed by more general mechanisms than something like face or rug (eccentric stimuli)

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

How do eccentric and ordinary stimulus domains differ?

Eccentric domains may require specialized computational mechanisms, while ordinary categorization may use general mechanisms shared across domains

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

What does Fodor mean when he says categorizing cows may work like categorizing rugs?

Both may use the same general categorization mechanism, rather than separate specialized modules

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

A theory for similar categorization of general stimuli (not eccentric)

Prototype Theory

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Prototype Theory

Proposes that people organize objects and concepts around a central, ideal example (prototype) rather than using strict rules

Ex. Is a German Shepard or Pomeranian more a typical ex for a dog?

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Prototype

Mental representation of features of category (EARS, FUR, PAWS, etc) + weights for typicality → if an object is similar to the prototype for X, then categorize it as an X

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

What is typicality more privileged in for prototype theory?

In speed, accuracy, and development

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Fodor Claim on Categorization Mechanism

Fodor suggests that we can use the same general mental process to categorize many different objects; The categories are different, but the underlying process (comparing an object with what you know about a category) might be the same

We don't necessarily need a separate mental module for every category of object

Ex. Cow → we recog features of cow

Ex. Dog → we recog features of cow

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Sentences + Categorization for Fodor

Sentences are too eccentric so they need a specialized categorization mechanism

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Is categorizing cows a distinct computational task from categorizing rugs for ppl?

No

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Why might recognizing other cows require a specialized mechanism for cows?

Recognizing conspecifics (members of the same species) can involve specialized tasks, such as identifying potential mates or herd members

Ex. Cows categorizing cows would be them categorizing conspecifics whereas if ppl do it its the same as categorizing cows

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

What might one + the same domain need for categorization?

One + the same stim domain may need specialized mechanisms for some purposes + not others

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Fodor on how we can infer domain-specific mechanisms

Fodor thinks we can look at the stim domain + how “eccentric” it is + infer domain specific mechanisms

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Issue with Fodor on how we can infer domain-specific mechanisms

Computational task is more important

Ex. For us → categorizing cows + rugs = same sort of task (aka recog obj)

Ex. For cows → categorizing cows = recognizing conspecifics (more specific computational task)

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Phoneme Perception

A sound may be heard as a consonant in speech but when isolated may hear a whistle or glide → what matters is whether the system is trying to compute the function (figure out what the sound is vs figure out what the phoneme is)

Ex as to how domain specificity depends on comp task

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Inversion Effect

Processing faces involves processing arrangement of parts rather than each indiv part alone; Effect is species specific (chimps show stronger inversion effect for conspecifics than humans)

Ex of how domain specificity depends on comp task

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Inversion Effect + Chimps

Effect is species specific; Chimps show stronger inversion effect for conspecifics than humans

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Does Fodor care about dif b/w individuals?

No

Domain Specificity + 9 Characteristics of Modules (some modules may lack some features)

Mandatoriness

Input systems are not under “executive control” → you can choose what to think about but you CANT help but parse a sentence or see whats in front of you; Input systems operate automatically when the relevant information is available to them

You can control your thoughts but you cant unsee a sentence (ants in a sentence pattern for ex) or not parse a sentence you hear

9 Characteristics of Modules (some modules may lack some features)

Stroop Task

Must say color of word; It takes more effort to not say red (if the word says red) bc you still process it as a word → evidence that lexical identification is automatic/mandatory

We would be much better at the task if we could not read the word

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

What does Stroop Task show evidence of?

Evidence that lexical identification is automatic/mandatory

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

2 Caveats/Qualification for Mandatoriness from Fodor

1. Attention

2. Re-Interpretation

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

Attention

If youre really absorbed in some task (eg. game), you may fail to parse what someone says; When inputs are available, the operation of the parser (transforming inputs into outputs) is still mandatory

Fodor → this minimizes the INPUT to the parser, or the extent to which the OUTPUT is recognized

2 Caveats/Qualification for Mandatoriness from Fodor

Re-Interpretation

Painters (or experts) may be able to “see” the 2D projection of 3D objects (an obj)

Fodor → This is a post-module re-interpretation of the output, NOT an effect on processing within the module/knowledge (a cog thing)

2 Caveats/Qualification for Mandatoriness from Fodor

When is operation of parser mandatory?

When inputs are available

2 Caveats/Qualification for Mandatoriness from Fodor

Do input systems operate auto when the relev info is available to them?

Yes

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

Try hearing ‘All Gaul is divided into three parts’ as an acoustic object Example for Mandatoriness

Fodor is asking you to compare two ways of hearing the exact same sentence.

• Normal hearing → You automatically hear and understand the words and their meaning.

• Hearing it as an acoustic object → You try to focus only on the sounds, as though the sentence were just a collection of noises rather than meaningful language.

Second is difficult because your brain automatically processes the sounds as language → Shows Mandatoriness bc our language-processing system automatically analyzes speech when the input is available

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

Semantic Satiation

When you repeat a word over + over again so it loses its meaning; A potential counterex to mandatoriness

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

Fodor on Processing in Central Cog

Processing in central cog is NOT as mandatory as input systems (to counterex of semantic satiation) → There are some things in central cog that are NOT under our control but even so, visual perception is even LESS in control

Mandatoriness + 9 Characteristics of Modules (some modules may lack some features)

Counterex to Fodor’s Idea of Processing in Central Cog

ToM, Implicit Bias

New Look Psych

Perception is driven by expectations + values; Ppl are slower + less accurate in recog the color of a playing card if it doesnt match the suit

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Card Study Def

Ppl are slower + less accurate in recog the color of a playing card if it doesnt match the suit → shows information encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Card Study + Result

Shows information encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Coin Study Def (Bruner + Goodman)

Coins look larger to kids living in poverty than to wealthy children; Challenges info encapsulation bc if ppl’s values or needs influence their visual perception, then perception may not be completely insulated from other information

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Coin Study + Info Encapsulation (Bruner + Goodman)

Challenges info encapsulation bc if ppl’s values or needs influence their visual perception, then perception may not be completely insulated from other information

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Fodor on Input System Access to Perception

We dont want input systems to have access to all of your expectations + values → quicker to avoid having to sift through all of the background info

Info encapsulated systems avoid this prob

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

What prob do info encapsulated systems avoid?

Having to sort through all of background info

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

What does it mean when modules are encapsulated?

Modules only have access to a proprietary store of info (specialized info that does NOT include all of your background knowledge, feelings, etc)

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Do modules have knowledge about central cognition?

No

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Light from Above Prior

Explained by information encapsulation; Bc visual input is ambig, brain default assumes that light comes from overhead → you know that a raised footprint is the same reversed image of an unraised footprint + the raised one is just an artificial light source but the vision system doesnt know that

Aka you may believe that in the room youre in, the lighting is unusually coming from below but you keep seeing the left as convent + right as concave

Your visual system doesnt have access to your knowledge → only the light from above prior

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Can chickens raised with light from below update their prior?

No

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

McGurk Effect + Info Encapsulation

Effect keeps working even after you know he’s saying ‘ba’ because it is sep from knowledge

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Muller-Lyer Illusion + Info Encapsulation

The two lines with arrows illusion that are the same length but top one looks longer →You understand that they arent changing lengths but the visual system CANT access that info even tho its relevant

Evidence for info encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Checkerboard Illusion + Info Encapsulation

We see A + B tiles as dif colors bc out visual systems are sensitive to shadows )making one under the cylinder seem darker + the one outside of the shadow seem lighter) → Even tho we know that theyre the same color we CANT see it that way

The visual system cant access that info that it is the same color → More evidence of info encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Grid Illusion + Info Encapsulation

See black dots on a grid but they disappear when we look at them/try to find them → Bc of the way visual system processes peripheral info + black/white

Visual system making smart assumptions but cant change it; More evidence for info encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Curved Line Illusion + Info Encapsulation

Think you see curved lines on a grid but there arent any → Doesnt go away even when you know

More evidence for info encapsulation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Motion Illusion + Info Encapsulation

Think circle is moving when it is not; Cannot unsee even when you know it isnt

More evidence for info encapsulating

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Color Illusion + Info Encapsulation

The blue + green are the same in the image; This is an inference that the visual system made + cannot be unseen even with knowledge

More evidence for info encapsulating → The visual system cant access central cog info

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Illusions are immune to what?

Cognitive influence; They still work even when you know how they work

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Hypothesis as to why Muller-Lyer Illusion works Name

Carpentered World Hypothesis

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Carpentered World Hypothesis for Muller-Lyer Illusion

Longer lines look farther away bc of corners constructed in certain cultural environments; There are lots of right angles in a carpentered world

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Blind Kids + Muller-Lyer Illusion

Blind kids that have cataracts removed are immediately susceptible to Muller-Lyer illusion → goes against Carpentered World Hypothesis

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

African Kids + Muller Lyer Illusion

Kids in cultures that do not have carpentered world are LESS susceptible to Muller-Lyer illusion → evidence for Carpentered World Hypothesis

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

What does visual system seem to rely on?

Info about length cues that are innately available to vision → remains impossible to unsee

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

What is vision insensitive to?

Expectations/beliefs; Its stubborn

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Perception

Cognitively impenetrable + stubborn → not affected by central cognition + knowledge, etc

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Point of Perception/Why Encapsulation? (Fodor)

To contradict belief; You dont want your visual system to tell you what you EXPECT to be out there, you want it to tell you what IS out there, even or esp when it defies expectation

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

What does perception have to perform?

Quick, unconscious, inferences

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Encapsulation

Constraints on info that can be used as a premise; Explains why input systems are FAST (since you can ignore most of the info stored in mind + dont need to search through all of your knowledge to see what is relevant)

Input systems make reliable assumptions

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

How does Encapsulation explain why input systems are fast?

Bc you can simply ignore most of the info stored in your mind; No need to search through all of your knowledge to see what is relevant → input systems just make reliable assumptions

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

3 Caveats about Encapsulation

1. Perceptual outputs can be corrected by background knowledge after perception (caveat A)

2. Cognitive influence on a similar mechanism doesn’t prove the input system itself is penetrable (caveat B)

3. Top-down processing within an input system can occur without violating encapsulation (caveat C)

Informational Encapsulation + 9 Characteristics of Modules (some modules may lack some features)

Caveat A about Encapsulation

Fodor identifies perception with the fixation of belief → you see the lines in the Muller-Lyer illusion as 2 dif lengths but you know they are the same length

Visual system produces the perceptual output, and your broader cognition can correct it afterward

3 Caveats about Encapsulation

Caveat B about Encapsulation

If the mind is breaking part into dif systems that make its own inferences → it is possible there is redundancy, aka mult systems performing the same function

Context might help you recognize a missing word in a sentence. But that could be central cognition filling in the gap, rather than the language input system itself being influenced by context

Fodor’s ex → what word comes next in ‘i have the pepper, but please pass the…’ → if the signal is degraded (cant hear it well) ppl will understand the sentence better if it ends in ‘salt’ than ‘dog’ (high vs low cloze)

Fodor’s Reply → parsing a sentence is not done SOLEY by an input system, you can also cog figure out what the sentence should be

3 Caveats about Encapsulation

Fodor’s Ex + Reply for Caveat B about Encapsulation

Fodor’s ex → what word comes next in ‘i have the pepper, but please pass the…’ → if the signal is degraded (cant hear it well) ppl will understand the sentence better if it ends in ‘salt’ than ‘dog’ (high vs low cloze)

Fodor’s Reply → parsing a sentence is not done SOLEY by an input system, you can also cog figure out what the sentence should be

High vs Low Cloze

What word comes next in ‘i have the pepper, but please pass the…’ → if the signal is degraded (cant hear it well) ppl will understand the sentence better if it ends in ‘salt’ than ‘dog’ (high vs low cloze)

No effect on high vs low cloze when words are clear + presented rapidly

3 Caveats about Encapsulation

Effect on High vs Low Cloze when words are clear + presented rapidly

None

3 Caveats about Encapsulation

Caveat C about Encapsulation

There can be “top-down” info flow WITHIN input systems WITHOUT violating encapsulation

Fodor’s ex → if the ‘s’ in ‘legislation’ is replaced with a cough, ppl can still hear the ‘s’ → must be a top-down effect, info about the word identity affecting phoneme perception

Fodor says this is NOT a violation of info encapsulation bc it goes underneath the perceptual system according to Fodor, so there are still no changes from central cog

Another ex → Obj Perception Warps Spatial Perception, Lexical Priming

3 Caveats about Encapsulation

Fodor’s Example + Explanation for Caveat C

Fodor’s ex → if the ‘s’ in ‘legislation’ is replaced with a cough, ppl can still hear the ‘s’ → must be a top-down effect, info about the word identity affecting phoneme perception

Fodor says this is NOT a violation of info encapsulation bc it goes underneath the perceptual system according to Fodor, so there are still no changes from central cog

3 Caveats about Encapsulation

Object Perception Ex Caveat C

Obj perception warps spatial perception; In a black box with 2 red dots and a box without the black box + 2 red dots, the one with the black box seems closer together

3 Caveats about Encapsulation

Lexical Priming + Caveat C 2 Possibilities

‘I have the pepper, but please pass the…”

1. 1 possibility → you reason that ‘salt’ is the most likely option

2. Another possibility → your input system for lexical identification stores associations b/w words (pepper + salt have co-occured before + your lexical input registered this as a co-occurance by forming a strong association, so pepper activates salt)

This priming LOOKS smart bc its sensitive to semantic relations but Fodor argues its dumb + encapsulated (just an associative network that registers freq) → association mimics context sensitive knowledge when it is brute + simple

3 Caveats about Encapsulation

What does Lexical Priming look like + what does Fodor argue it actually is? Caveat C

This priming LOOKS smart bc its sensitive to semantic relations but Fodor argues its dumb + encapsulated (just an associative network that registers freq)

Association mimics context-sensitive knowledge (if she said she has the pepper, then she may want the salt bc salt + pepper are complementary spices) but it is brute + simple (salt ←→ pepper)

3 Caveats about Encapsulation