Category Clustering and Morphological Learning

Abstract

Inflectional affixes expressing the same grammatical category (e.g., subject agreement) tend to appear in the same morphological position in the word. We hypothesize that this cross-linguistic tendency toward category clustering is at least partly the result of a learning bias, which facilitates the transmission of morphology from one generation to the next if each inflectional category has a consistent morphological position. We test this in an online artificial language experiment, teaching adult English speakers a miniature language consisting of noun stems representing shapes and suffixes representing the color and number features of each shape. In one experimental condition, each suffix category has a fixed position, with color in the first position and number in the second position. In a second condition, each specific combination of suffixes has a fixed order, but some combinations have color in the first position, and some have number in the first position. In a third condition, suffixes are randomly ordered on each presentation. While the language in the first condition is consistent with the category clustering principle, those in the other conditions are not. Our results indicate that category clustering of inflectional affixes facilitates morphological learning, at least in adult English speakers. Moreover, we found that languages that violate category clustering but still follow fixed affix ordering patterns are more learnable than languages with random ordering. Altogether, our results provide evidence for individual biases toward category clustering; we suggest that this bias may play a causal role in shaping the typological regularities in affix order we find in natural language.

Keywords: Artificial language learning; Category learning; Learning biases; Morphology.

Fig 1

Examples of shapes and names presented in the experiment.

Fig 2

Participants’ scores on suffix test trials aggregated across the three test phases and grouped by grammar. Shaded dots represent individual scores, and circled dots depict the groups’ means and standard errors.

Fig 3

Mean participant accuracy score by grammar and test block. Left: Empirical data obtained from the experiments. Shaded circles represent individual scores, and larger circles represent more individuals. Filled circles represent mean accuracy scores, and the error bars represent the standard error. At Block 3, the mean is calculated across both novel and familiar items together; the mean scores split by novel versus familiar items are additionally illustrated by the upward (seen items) and downward (unseen “novel items”) triangles. Right: Model estimates of the fitted Bayesian beta‐binomial model for the fixed effect of the block. Thick lines represent the predicted accuracy means conditioned on grammar and block. The shaded area shows the 95% uncertainty intervals.

Fig 4

Number of training views. Left: Empirical data obtained from the experiments. Shaded circles represent individual scores, and larger circles represent more individuals. Filled circles represent mean accuracy scores, and the error bars represent the standard error. Right: Model estimates of the fitted Bayesian Beta binomial model for the training views fixed effect. Thick lines represent the predicted accuracy means conditioned on experimental condition and block. The shaded area shows the 95% uncertainty intervals.

Fig 5

Model's posterior distribution densities for all fixed effects and their interactions, along with their point mean estimates (solid black line) and 95% uncertainty intervals (dashed gray lines).

Fig 6

Participants’ scores on stem test trials aggregated across the three test phases and grouped by grammar type. Shaded dots represent individual scores, and circled dots depict the groups’ means and standard errors.