Annika McDermott-Hinman, Rowena Garcia, Scott Anderbois and Roman Feiman
Affiliation: Brown University, Leibniz-Centre General Linguistics, Brown University, Brown University
Category: Psychology
Keywords: Language acquisition, Negation, Conceptual development, Cross-linguistic comparison
Date: Thursday 4th of September
Time: 17:00
Location: GSSR Plenary Hall (268)
View the full session: Cognitive Development
Any child learning language must learn the mapping from words in their input to concepts in their mind. Thus, by the time that a child has learned that the word “not” means negation, they must have the abstract, logical concept of negation. But how did such a concept get into the child’s head prior to that point? We explored whether the process of learning the corresponding language itself contributes to children’s construction of abstract, logical concepts like negation.
Children learn the language for negation slowly over several years, and their earliest uses of negative language are limited to particular social functions, like rejection and prohibition (Bloom, 1970). Meanwhile, they don’t reliably use or understand negation words to mean logical negation until after their second birthdays (Feiman et al., 2017). This protracted acquisition of negation words may reflect merely children’s difficulty in identifying which word in their language means negation—purely a language learning problem. However, children may also be slow to learn the meaning of negation words because they lack the logical concept of negation; conceptual development may be required, in addition to language learning. If conceptual development is required, then the structure of the language the child is learning may help to scaffold it.
We investigated this possibility by leveraging cross-linguistic differences in the expression of negative social communicative functions. In some languages, the words that express functions such as rejection and prohibition are also the words that express logical negation—for example, “no” in English can deny the truth of an utterance (“Is Pluto gaseous?”; “No!”), but it can also stand alone to reject an offer (“Do you want a cookie?”; “No!”) or prohibit an action (“Can I go to the park?”; “No!”). In other languages, like Hebrew, there are specific words to express these functions (e.g., prohibition is expressed in Hebrew using אל (‘al’)). While logical negation words can be used in those languages to express these functions (e.g., negation composing with “allowed” to express prohibition), the presence of a lexical alternative makes them less likely to do so. Thus, while a child learning English may hypothesize that the logical negator “no” means prohibition (as in “No climbing!”), a child learning Hebrew is unlikely to form a similar hypothesis about Hebrew’s logical negator, לא (‘lo’).
If young children do not initially have the concept of negation, then early mappings between negation words and more specific negation-related functions, such as prohibition and rejection, may help to scaffold concept learning (Pea, 1980). A child learning English, who has already generated the separate hypotheses that the word “no” means rejection and prohibition and nonexistence, may have an easier time abstracting from this grouping the common concept of negation, as compared to a child learning Hebrew, in which prohibition (אל (‘al’)) and nonexistence (אין (‘eyn’)) are expressed by separate words. On the other hand, if the concept of negation is already available to children by the time they are learning language, then it may be easier to learn a language like Hebrew, in which children need to map fewer separate meanings onto a single negation word.
We tested whether children learning languages like English, with many negative functions clustered on a single negation word, learn logical negation faster than children learning more differentiated languages like Hebrew. We compared the acquisition of negation words across four languages—Spanish, English, German, and Hebrew—in which negation words can express varying numbers of specific and early-acquired negative social functions (Table 1).
We examined negation production of children ages 16-30 months in transcripts drawn from CHILDES (MacWhinney, 2000). Transcripts in each language consisted of ~40,000-130,000 utterances, containing ~4,000-7000 negations, which were coded by two native speakers. Coders were trained to categorize negative utterances by their meanings according to a schema developed by Bloom (1970), identifying negations as expressing rejection (including prohibition meanings), nonexistence, or (logical) denial negation.
To determine whether the rate of acquisition of logical denial negation differs, relative to overall language acquisition, across languages, we fitted a Bayesian logistic regression model predicting whether each child utterance expressed denial. We took the mean length of children’s utterances (MLU) as a measure of their overall language acquisition. To ensure that this comparison was valid across languages, we scaled children’s MLU by mothers’ MLU in each language. The model included Language, scaled child MLU, and their interaction, with a random slope and intercept for each child.
We found that, compared to English, children produced overall more denials in German (logOR 1.08, 95% CI [.66, 1.52]) and Spanish (logOR .58, 95% CI [.22, .96]). Denial production in Hebrew was slightly higher than in English, though this effect was not statistically robust (logOR .23, 95% CI [-.25, .69]). Children producing more denials in German and Hebrew than in English is consistent with children learning denial earlier in languages with fewer specific meanings mapped to a logical negation word. However, in such a world, we would predict Spanish and English to pattern together, and we found instead that Spanish-speaking children produced more denial than English-speaking children. Further, in German logical negation words express more specific meanings than in Hebrew (Table 1), and so we would also expect children to learn denial later in German than Hebrew. However, children produced fewer denials in Hebrew than in German (logOR -.84, 95% CI [-1.42, -28]; Figure 1). We found no differences across languages in the rate of increase in denial production over MLU. These findings suggest that differences in denial production are more likely due to elements of word learning that are unique to each language, rather than to a boost to conceptual development from the clustering of negative meanings on a single word. For instance, children learning Spanish have only one negation word to learn, which may facilitate the task of mapping meanings to words. Children learning German are uniquely exposed to a negative existential quantifier (“kein”) that can express both nonexistence and logical negation. Future work will explore whether this confluence might facilitate the language-learning task for German-speaking children.