UCI Experts On: Language
“When we know a language, we know so much more than we can put into words. As a language
scientist, part of my job is identifying and understanding that implicit knowledge,
unpacking what it is that we know when we know a language. I'm Greg Scontras, associate
professor of language science here at UC Irvine, where I direct the Meaning Lab.
One area of linguistic knowledge that I've studied is how speakers order multiple adjectives.
For example, big brown box sounds much better to me than brown big box. And if you're a speaker of English, I bet you'll agree this isn't random. English speakers have consistent preferences, even if we can't explain them. What's fascinating is that we find the very same preferences in unrelated languages, for example, in Tagalog, the national language of the Philippines.
Since these languages evolved independently, it's likely that the preferences are coming from universal aspects of the human mind. It gets even more interesting.
Some languages are like English or Tagalog. Adjectives come before the nouns they modify, as in big brown box, but some languages place adjectives after the nouns they modify, like in Arabic or Vietnamese. In those languages, we find the mirror image of the English like preferences, so speakers prefer box brown big over box big brown. These mirror image preferences suggest that a full explanation can't rely just on linear precedence, since the adjectives that speakers prefer to come early in English, Arabic speakers prefer to come late.
At issue is the relative distance of adjectives from the nouns they're modifying. The first step to an explanation is careful documentation of the preferences to be explained. In my lab, we've used large scale web-based paradigms to elicit preferences from speakers of different languages. When you get lots of judgments from lots of speakers for lots of adjectives and nouns, a picture emerges of the general patterns people prefer.
It's been especially rewarding to team up with student speakers of different languages. The students get involved in language science research, leading the data collection for the languages they speak as a way of corroborating and expanding the data we gather in our controlled elicitation paradigms. We also analyze mountains of text from dozens of languages. The text comes from things people have said on the internet, web articles, social media, television transcripts - you name it.
Data science tools allow us to carefully comb through all the text to document the ordering regularities.
With a handle on what the preferences are across linguistically, the task then turns to explaining them. One thing we can say with certainty is that there's no single factor that explains ordering preferences. There are low level factors like word length, with shorter adjectives preferred before longer ones. We also find that adjectives that often modify specific nouns are preferred closer to those nouns.
But the factor I focus on in my work concerns a specific aspect of adjectives, meaning how subjective the property is that the adjective describes. What we found in language after language is that less subjective adjectives are preferred closer to the noun. Consider the adjective big. You and I might disagree about whether a box is big, and we might disagree without either one of us being wrong.
Big is highly subjective, but contrast that with brown. You and I probably agree on what counts as brown and what doesn't. Brown has relatively low subjectivity, and in both English-like languages and Arabic-like languages, speakers have a strong preference for less subjective brown to occur closer to the noun than more subjective big. The explanation lies in the way that meanings get built up compositionally, hierarchically outward from the noun.
For example, suppose you start with a noun – box - and there are lots of boxes, but you want to refer to a specific one. You'll need to give me more information about which box you intend. You can restrict the possible boxes with an adjective like brown, so that now you're only talking about the brown boxes, and you can restrict that set of brown boxes with another adjective, say big, so that you can pick out the big brown box that you're talking about.
The trick here is that you want to use adjectives with low subjectivity earlier in this incremental restriction process, in order to avoid alignment errors between speakers and listeners. In other words, placing adjectives with low subjectivity closer to the noun gives you a better chance of successfully communicating with your listener. This reasoning applies the same in English-like languages and Arabic-like languages.
In either case, subjectivity-based ordering preferences come from a pressure toward successful communication. We find the same preferences in language after language, because language is being used by its speakers to communicate about the world around us.
Computer simulations support this explanation. Across hundreds of thousands of simulated conversations, we find that adjective orders that place less subjective adjectives closer to the noun lead to more successful communication. So that is why you prefer big brown box over brown big box. You probably wouldn't have been able to explain it until I pointed it out to you, but you already had this knowledge by virtue of knowing how to speak English.
The seemingly simple preference arises from the fact that language has been shaped in a way to support successful communication, and the communicative pressures that are shaping our ordering preferences, are likely to apply broadly across all areas of language. Sequences of multiple adjectives just happen to be a convenient domain in which to observe those pressures at play.”