LANG209: Animals and Language

Animals and Language

Introduction to Animal Communication and Language

• Core Questions: The LING209 course explores two fundamental questions:

  • How does human language relate to other animal communication systems?
  • Can animals learn human-like language abilities?

• Agenda: The lecture is structured into three main topics:

  1. Animals and grammar.
  2. Primate language learning: words.
  3. Primate language learning: syntax.

Animals and Grammar

Vervet Monkey Vocabulary

• Vervet monkeys demonstrate a specialized vocabulary for alarm calls.
• They possess three distinct alarm calls, each corresponding to a specific predator:
  • Eagle: Prompts looking up or seeking aerial cover.
  • Snake: Prompts looking down or standing on hind legs to search for the snake.
  • Leopard: Prompts climbing a tree or seeking refuge in branches.
• These calls elicit appropriate, distinct behaviors, suggesting referential communication (the calls refer to specific entities in the environment).

Bee "Language"

• Honeybees utilize complex dances to communicate information about food sources, often referred to as a "language."
• Round Dance: Used for food sources located nearby.
  • The vigor of the dance indicates "how much" food is available.
• Waggle Dance: Used for food sources located at a distance.
  • The vigor of the dance indicates "how much" food.
  • The speed of the dance indicates "how far" the food is.
  • The angle of the dance (relative to the sun or gravity) indicates "which direction" the food is.
• Question for Reflection: Do bees have "language"?
  • While they communicate novel messages (specific location and quantity of food), their communication operates within a limited, not infinite, domain of information (e.g., only about food, not abstract concepts, past events, or future plans).

Starling Syntax

• European starlings exhibit complex vocalizations, raising questions about avian "syntax."
• Song Variety: There is a wide variety of song patterns across different starling species.
• Learned vs. Innate: Some aspects of their songs are learned, similar to humans, while others are innate, like sparrow songs.
• Complexity: Starling songs can consist of long, intricate sequences.
• Internal Structure: These sequences often possess an internal structure, referred to as "motifs."
• Holistic Meaning: A key distinction from human language (and even bee communication) is that the meaning of starling song is holistic; it is not generated from meaningful individual pieces. This means individual notes or motifs do not carry independent meaning that combine compositionally to form a larger, complex message.
• Reference: Gentner et al. (2010 Nature) is cited for research on starling song.
• Question for Reflection: Do starlings have "language"?
  • They vocalize complex, systematic patterns. However, they do not appear to use these patterns to generate novel meanings in a compositional way, unlike human language.

Chimpanzee Grammar: Bonobos and Compositionality

• A recent study by Berthet et al. (2025 Science) investigated "extensive compositionality in the vocal system of bonobos."
• Methodology: Statistical analysis of the contexts of different vocal calls was used.
• Key Finding: Vocal combinations modify the usage patterns (and potentially the meaning) of individual calls.
• Compositionality Defined:
  • The capacity to combine meaningful elements into larger meaningful structures.
  • A hallmark of human language, where meaning is determined by the meaning of the parts and their combination.
  • Trivial (or Intersective) Compositionality: Each element independently contributes to the meaning of the whole, and the combination is interpreted by the conjunction of its parts (e.g., "blond dancer" refers to a person who is both blond AND a dancer).
  • Nontrivial (or Nonintersective) Compositionality: Units do not contribute independent meaning; one part of the combination modifies the meaning of the other (e.g., "bad dancer" refers to a dancer who is bad at dancing, not a bad person who also happens to be a dancer).
• Significance for Bonobos: The study found that not only does each call type occur in at least one compositional combination, but three of these compositional combinations also exhibit nontrivial compositionality.
• Implication: These findings suggest that compositionality is a prominent feature of the bonobo vocal system, revealing stronger parallels with human language than previously thought, addressing a lack of unambiguous evidence for nontrivial compositionality in animals from systematically collected quantitative data.
• Semantic Space Analysis: Researchers adapted a distributional semantic framework.
  • Distributional Hypothesis: Words with similar meanings are used in similar contexts.
  • Method: Meaning represented in a semantic space by converting each word into a vector, quantifying relationships using geometric relations.
  • Statistical Tool: Multiple Correspondence Analysis (MCA) was performed on categorical data (utterance types and associated contexts, or Features of Contexts - FoCs).
  • Results: The first five dimensions of the MCA explained approximately $24<br /> brace$ of the variance in the data ($15.00 + 3.12 + 2.48 + 2.11 + 2.14<br /> brace$).

Take-aways on Animals and Grammar

• Several non-human animals demonstrate communication systems based on complex, generative patterns (e.g., bee "waggle" dance, starling song).
• However, these systems do not approach the communicative power of human language due to limitations such as:
  • Absence of semantics (meaningful components) in birdsong.
  • Limited domain of meanings for bees (e.g., only food-related information).

Primate Language Learning: Words

Kiki (Chimpanzee)

• Background: Kiki was a chimpanzee raised alongside a human infant and exposed to spoken English.
• Speech Production: Kiki could not produce human speech sounds.
• Reason: This inability is attributed to fundamental differences in the anatomy of the vocal tract between chimpanzees and humans.

Koko (Lowland Gorilla)

• Language Modality: Koko was taught American Sign Language (ASL).
• Vocabulary Size:
  • Production: Estimated to produce approximately $1,000$ signs.
  • Comprehension: May comprehend around $2,000$ signs.

Kanzi (Bonobo Chimpanzee)

• Learning Context: Kanzi initially learned symbols by observing his mother, who was being taught symbols.
• Communication Method: Kanzi uses "lexigrams" via a talking keyboard to indicate nouns and verbs.
• Sentence Comprehension: May comprehend simple sentences.
• Demonstration of Creativity: While on a walk, Kanzi spontaneously signed "water" and "bird" in combination to refer to a "duck." Researchers had not explicitly taught him the compound "water-bird" for duck, suggesting a novel conceptual combination.
• Question for Reflection: Koko and Kanzi show a facility with learning words. These skills:
  • Can be partly consistent with behaviorist accounts of ape language via association (learning word-object pairings).
  • However, Kanzi's creative combination for "duck" suggests a more cognitivist account involving internal mental representations and the ability to combine concepts in novel ways, going beyond simple stimulus-response associations.

Primate Language Learning: Syntax

Nim Chimsky (Chimpanzee)

• Background: Nim was a chimpanzee extensively exposed to American Sign Language, both in "uncontrolled" home settings and controlled laboratory sessions.
• Word Combinations: Nim produced extensive word combinations.
• Grammar Analysis: However, in-depth analysis of his productions suggested a very limited systematic grammar.
• Evidence of Limited Grammar:
  • Mean Length of Utterance (MLU): While children's MLU typically increases steadily with age, Nim's MLU plateaued, suggesting a lack of grammatical growth.
  • Lack of Systematic Word Order: Although Nim showed preferences for certain word orders (e.g., V(t) + me like "give me" was more frequent than me + V(t) like "me give"), these patterns were not systematic or consistent enough to be considered human-like grammatical rules.
  • Frequent Combinations: Examples of frequent two-sign combinations include "play me," "me Nim," "tickle me," "eat Nim," "more eat," "me eat," "Nim eat."
  • Lack of Compositional Meaning in Combinations: Many longer utterances appeared to be repetitive or permutations of simpler phrases rather than combinations with emergent new meanings.
• Question for Reflection: If Nim were to learn human-like language structures, we would expect to find:
  • Systematic word-order patterns that are consistent across different utterances and apply broadly, indicating the presence of grammatical rules, rather than just frequent co-occurrence.
  • While frequent use of two or more words together (A) was observed, and utterances often stayed in the "here and now" (C), the crucial missing element was (B) systematic grammatical patterns and productive syntax.

Take-aways on Primate Language Learning

• Great apes, when exposed to human language structures, are able to learn certain aspects well.
  • For instance, they can acquire a substantial vocabulary, as evidenced by Koko's large sign language vocabulary and Kanzi's lexigram usage.
• However, despite their capacity for word learning, great apes generally do not seem able to acquire or produce human-like grammar or complex syntax, which involves systematic rule-based combination of words to create novel meanings.