Perceptual Development: Narrowing, Categorical Perception, and Multimodal Integration

Follow-up on Causation and Infant Learning

• The lecture begins with a recap of a final study on how infants learn about cause and effect, specifically regarding the features of toy trucks and their predictive outcomes.

• Scenario Details: Infants were shown toys where specific features were predictive of functions in ways that do not make mechanistic sense to adults (e.g., the type of wheels predicting a sound from the top, or the top part predicting if the wheels would turn).

• Observations of Violation: Infants were shown correlations repeatedly, then shown new examples that violated these patterns. Surprise or longer looking times indicate that they had learned the initial correlations.

• Findings by Age Group:

  • $14$-month-olds: Showed evidence of learning these "weird" statistical regularities. They soak up co-occurrences without considering if they make mechanistic sense.

  • $18$-month-olds: Did not learn these correlations. They failed to pick up on the statistical regularities because the relationships were "weird" and scientifically implausible.

• Theoretical Interpretation: $18$-month-olds have a more refined understanding or a "mechanistic bias" that prevents them from learning arbitrary or non-sensical correlations. This is compared to the "gravity bias," where older children misapply a correct principle (objects drop in a straight line) to specific instances where it may not apply.

Introduction to Perceptual Development

• Target Reading: Chapter $6$, specifically section $6.2$, is the most relevant for today's material.

• Lecture Objectives:

  • Identify misconceptions regarding how perception functions.

  • Define and provide examples of Categorical Perception.

  • Define and provide examples of Perceptual Narrowing.

  • Explain Multimodal Integration and its relationship to perceptual narrowing.

Misconceptions: The Camera and Microphone Metaphor

• A common misconception is that we perceive the world exactly as it is (objective rendering).

• The Camera Metaphor: The belief that visual perception is passive, like opening a lens and processing light into an objective image.

• Counter-examples (Visual Illusions):

  • Panel A and Panel B: Shading and shadow information cause the visual system to perceive two identical shades as different. Pixels are objective to a camera, but context-dependent to the human eye.

  • Contextual Rectangles: Identical orange rectangles (same RGB values) appear to be different colors when placed in a complex visual array.

  • Shepard Tables: Rectangular table tops that are identical in pixel dimensions and shape appear to be different sizes/lengths due to 3D orientation and contextual cues like table legs.

  • Ebbinghaus Illusion: Identical orange circles appear to be different sizes depending on the size of the surrounding circles.

  • Peripheral Drift Illusion (Rotating Snakes): A completely static image that creates the sensation of movement because the visual system is primed for exploration and motion.

  • "The Dress": A famous internet example where individuals perceive the same stimulus as either black and blue or white and gold. This illustrates that personal experience and internal construction vary even when the light hitting the eye is identical.

• The Microphone Metaphor: The belief that auditory perception is a literal recording of sound. This is debunked by examples where noise (like R2-D2 sounds) is uninterpretable until the listener knows the semantic meaning (e.g., "The Constitution Center is at the next stop"), after which the brain "cannot help" but hear the message.

Categorical Perception

• Definition: The tendency to group incoming sensory information that exists along a physical continuum into discrete, distinct categories.

• Example - Light Intensity: Light intensity is a continuous variable. Perception generally maps onto this reality accurately; we experience a smooth gradient from dark to light.

• Example - Hue/Color: Hue is also continuous, yet we perceive it categorically. We see a "jump" from one category (blue) to another (green) rather than a smooth shift, pulling our experiences toward specific category boundaries.

• Example - Speech Phonemes: Phonemes (basic units like $ah$, $ba$, $th$) are perceived categorically. Even when sounds are computer-generated across a continuum from $ba$ to $pa$, humans hear them as clearly and distinctly one or the other, rather than an intermediate sound.

The Development of Face Perception

• Babies show a preference for faces within hours of birth, leading to debates about whether face perception is pre-wired (innate).

• Comparative Evidence (Innate Wiring):

  • Tinbergen ($1950$s): Barnyard chicks with no prior exposure to predators freaked out when a silhouette moved in a direction indicative of a hawk, but not when it moved like a non-threatening goose. This suggests pre-wired threat detection.

  • Pecking Behavior: Newborn chicks prefer pecking at natural, round seed shapes over sharp-edged, artificial shapes.

• Human Infant Evidence (Johnson and colleagues):

  • Used paddles with different configurations (face-like, scrambled, or blank).

  • Found no innate preference for human faces specifically, but a preference for face-like configurations (defined as two dark elements at the top, one at the bottom).

• Theoretical Debates:

  • Nativists: Argue this evidence proves built-in biological knowledge.

  • Empiricists: Argue this can be learned through very early experience or even in utero (e.g., a motor map of one's own face).

• Developmental Timeline:

  • $3$ months: Preference for well-proportioned faces and familiar face types (the "familiar faces effect," often mislabeled as the own-race effect).

  • $6$ months: Infants can discriminate between different human faces and different monkey faces equally well (Generalists).

  • $9$ months: Ability to discriminate human faces improves, while the ability to discriminate monkey faces declines.

  • Sheep Face Study: $4$ to $6$-month-olds discriminate sheep faces well, but this ability is lost by $9$ to $11$ months.

Perceptual Narrowing

• Definition: A developmental process where the ability to perceive stimuli frequently encountered in the environment improves, while the ability to perceive stimuli not frequently encountered declines.

• This is described as an "expertise effect."

• Phonemic Narrowing (Patricia Kuhl):

  • $6$ months: Infants are "Citizens of the World" and can discriminate the phonemes of all languages ($40$ to $45$ in English, but many more globally).

  • $12$ months: Infants become "Culture-Bound Listeners." They lose the ability to detect contrasts in non-native languages.

  • Co-articulation: The way phonemes are produced depends on the surrounding sounds, making phonemic isolation a complex task that requires statistical learning from "Motherese" (or Parentese).

Multimodal Integration

• Definition: The process by which the brain combines information from different sensory modalities (e.g., vision, touch, sound) into a single, coherent perception.

• Integrating Vision and Touch (Meltsoff and Borton):

  • Infants were given a pacifier (dummy) that was either bumpy or smooth without being allowed to see it.

  • When shown pictures of both, infants looked longer at the image that matched their tactile experience (bumpy-to-bumpy or smooth-to-smooth).

• Integrating Sound and Vision (Lewkowicz and Ghazanfar):

  • Study Design: Newborns and older infants were shown videos of monkey faces making two vocalizations: a "coo" and a "grunt."

  • Newborn Results: In the no-sound condition, looking was at chance ($50$%). In the sound condition, they looked significantly longer at the face that corresponded to the audio (e.g., hearing a grunt made them look at the grunting face). This proves multimodal integration is present at birth.

  • $4$-month-old Results: The distinction between the sound and no-sound conditions gets smaller, and looking times move toward chance.

  • Connection to Perceptual Narrowing: As infants lose the ability to discriminate between individual monkey faces (due to lack of exposure), they simultaneously lose the ability to integrate those faces with their corresponding sounds. If the categories cannot be distinguished, they cannot be matched.

Questions & Discussion

• Question on Bilingualism: Does exposure to multiple languages prevent narrowing?

  • Response: Infants in bilingual households learn to maintain the discrimination of phoneme categories relevant to both languages. However, there may be some trade-offs in the "statistical peaks" of discrimination or fine-grained distinctions compared to monolinguals.

• Question on Dialects: Do the same principles apply?

  • Response: Yes, the same underlying statistical learning mechanism is likely at work, though the specific statistics of the input vary.

• Question on Motherese/High-Pitched Speech: Why do we speak that way to infants?

  • Response: One hypothesis is that the high pitch exaggerates sound distinctions, helping infants pick up on phonemic structures. The extent to which this is a universal biological trait versus a cultural practice is currently being researched by specialists at this university.

• Question on Surprisingness vs. Matching: Why do infants sometimes look longer at what matches (matching bias) and sometimes at what is new (surprisingness bias)?

  • Response: This is a major debate in developmental psychology. Generally, if an infant hasn't been habituated to a stimulus yet, they often show matching behavior; if they have been habituated, they show a novelty/surprise preference.