Generalization Predictions

The below sliders illustrate the pattern of generalization predicted as a function of the amount of data observed (N). Male denoted with hats. While it may appear that the model has learned the correct extension for the word with very few data points, this is not necessarily the case. For this visualization, the predictions are fixed to this family tree context. The model might actually make incorrect generalizations on other trees. Refer to Figure 2 in the main paper for how many data points are required for the model to learn the context independent "correct" meaning.

---

Pukapuka

Kainga

Based on the model, we would expect learners to initially over-generalize to all members of the learner's generation before restricting to just the men of the learner's generation.

Taina

Based on the model, we would expect learners to initially over-generalize to all members of the learner's generation and all women, then restricting to just the learner's generation before settling on just the women of the learner's generation.

Matua-tane

Based on the model, we would expect learners to initially over-generalize to all members of the learner's parent's generation, then restricting to just the men, then restricting to those related to the learner by blood.

Matua-wawine

Based on the model, we would expect learners to initially over-generalize to all members of the learner's parent's generation, then restricting to just the women, then restricting to those related to the learner by blood.

Tupuna-tane

Based on the model, we would expect learners to initially over-generalize to all men before restricting to members of the learner's grandparent's generation.

Tupuna-wawine

Based on the model, we would expect learners to initially over-generalize to all women before restricting to members of the learner's grandparent's generation.

---

English

Mother

Based on the model, we would expect learners to initially over-generalize to both parents before restricting to women.

Father

Based on the model, we would expect learners to initially over-generalize to both parents before restricting to men.

Brother

Based on the model, we would expect learners to initially over-generalize to their parent's children before restricting to men.

Sister

Based on the model, we would expect learners to initially over-generalize to their parent's children before restricting to women.

Grandma

Based on the model, we would expect learners to initially over-generalize to all blood related members of their grandparent's generation before restricting women.

Grandpa

Based on the model, we would expect learners to initially over-generalize to all blood related members of their grandparent's generation before restricting men.

Aunt

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to women, before exlcuding their parent.

Uncle

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to men, before exlcuding their parent.

Cousin

Based on the model, we would expect learners to initially over-generalize to all members of their generation before exlcuding their parent's children.

---

Turkish

Anne

Based on the model, we would expect learners to initially over-generalize to their parents before restricting to women.

Baba

Based on the model, we would expect learners to initially over-generalize to their parents before restricting to men.

Abi

Based on the model, we would expect learners to initially over-generalize to their parent's children before restricting to men.

Abla

Based on the model, we would expect learners to initially over-generalize to their parent's children before restricting to women.

Anneanne

Based on the model, we would expect learners to initially over-generalize to all blood related members of their grandparent's generation then restricting women, then to maternal relations.

Babaanne

Based on the model, we would expect learners to initially over-generalize to all blood related members of their grandparent's generation then restricting women, then to paternal relations.

Dede

Based on the model, we would expect learners to initially over-generalize to all blood related members of their grandparent's generation before restricting men.

Hala

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to women, before settling on their father's sisters.

Dayi

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to men, before settling on their mother's brothers.

Teyze

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to women, followed by maternal blood relations, before exlcuding their mother.

Amca

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to paternal blood relations, followed by men, before exlcuding their father.

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to men, before exlcuding parent's siblings and then parents.

Yenge

Based on the model, we would expect learners to initially over-generalize to all members of their parent's generation then restricting to women, before exlcuding blood relations and parents.

Kuzen

Based on the model, we would expect learners to initially over-generalize to all members of their generation before exlcuding their parent's children.

---

Yanomamö

Haya

Based on the model, we would expect learners to initially over-generalize to all blood related members of their parent's generation then restricting by coresidence, before restricting to men.

Naya

Based on the model, we would expect learners to initially over-generalize to all blood related members of their parent's generation then restricting by coresidence, before restricting to women.

Eiwa

Based on the model, we would expect learners to initially over-generalize to all members of their generation then restricting to men, before restricting to siblings and parallel cousins.

Amiwa

Based on the model, we would expect learners to initially over-generalize to all members of their generation then restricting to women, before restricting to siblings and parallel cousins.

Soaya

Based on the model, we would expect learners to initially over-generalize to all blood related members of their parent's generation then restricting to men, before narrowing to maternal brothers.

Yesiya

Based on the model, we would expect learners to initially over-generalize to all blood related members of their parent's generation then restricting to women, before narrowing to paternal sisters.

Soriwa

Based on the model, we would expect learners to initially over-generalize to all members of their generation then restricting to men, before restricting to cross-cousins.

Suaboya

Based on the model, we would expect learners to initially over-generalize to all members of their generation then restricting to women, before restricting to cross-cousins.

---

Pareto-Front