Leah Jumped the Morp Fracky: Early Findings in VerbCorner
Special thanks to Katherine Montas, Kayley Okst, Olivia Eastwood, Alexa Dudley and Lily Feinberg for writing this piece about an online study called VerbCorner that we run here in the lab!
Imagine this: you’re taking the SAT and get to the grammar section. You know A is the right answer, but you have no idea why. Often, we are following grammar rules without really knowing or being able to articulate what exactly they are. Read on to find out more!
VerbWhat?
Good question! Through the VerbCorner Project, professional scientists and citizen scientists are working together in an attempt to understand English grammar rules.
To understand VerbCorner, let’s consider the following: Some verbs make sense in some sentences, but not in others. For example, the sentence “Leah struck the vase” can be rephrased as “Leah struck at the vase.” In this case, the preposition “at” works well in the sentence. However, if we were to say, “Leah smashed at the vase” the sentence becomes ungrammatical. Why is this? That’s what VerbCorner is trying to figure out!
Oh! VerbCorner! But what’s the point?
Although native-born English speakers are extremely fluent and proficient in their language, nobody can actually list and explain all the rules of English grammar; most of the difficult stuff just comes naturally as a byproduct of language acquisition, based on what sounds right, or intuition. For example, if you are a native English speaker you know that “great green dragons” sounds better than “green great dragons”. Mark Forsyth’s The Elements of Eloquence explains that this is because English speakers unconsciously follow grammatical rules that they are completely unaware of.
Despite not being consciously aware of the rules, native English speakers do know which verbs go into which positions (also known as verb frames) in a sentence. For example, we all know that the verb amuse can be used in each of the following sentences:
- The movie amused the children.
- The movie’s special effects amused the children.
- Children amuse easily.
Amazingly, babies are able to learn the basic rules of how and when to use verbs in just a few short years! For this reason, scientists believe that babies do not learn rules for each verb and each verb frame on a verb-by-verb and frame-by-frame basis (since this would take forever!), but rather learn the patterns behind verb uses. However, it is very difficult to teach these patterns without explicitly knowing what they actually are. Figuring out the rules of grammar would enable us to teach English and address language difficulties in a more efficient way. The question is, what are these rules?
Though linguistic researchers have been working on this problem for a long time, we still don’t have the answer. The English language consists of numerous verbs and frames into which they can fit. The VerbCorner project is attempting to determine how the meaning of a verb and its place in the sentence are related to each other. To further explain this paradigm, think about how some verbs can take on different meanings depending on their context. For instance, the verb run has different meanings in each of the following sentences:
- She runs to the bath.
- She was running late to the bath.
- She runs a bath.
Researchers are exploring why context can change meaning so dramatically.
One of the current hypotheses, the Semantic Consistency Hypothesis, suggests that the meaning of a verb is important for determining what types of sentences the verb can fit in. Essentially, this hypothesis implies that there is a relationship between the meaning of the word and how it fits into different sentence frames. However, this is merely one hypothesis. Since previous investigations have only focused on a small number of verbs, researchers have yet to find conclusive results.
So, what’s the solution?
In collaboration with VerbNet, an online catalogue that groups 6,000+ verbs into different categories based on their possible uses and meanings, the VerbCorner project aims to confirm the validity of the listed syntactic frames and show a relationship between grammar and meaning. Data collection for this type of research is expected to take several years, despite using crowdsourcing (a method of collecting data from large numbers of people typically through the Internet). Because it will take so long, we have broken it down into multiple phases. Check out more about phase one below!
Explode on Contact?!
Research shows that humans respond more effectively to situations that can be related to real life and do not involve abstract scenarios. For this reason, in our “Explode on Contact” task, each concept (e.g., physical contact) was converted into a judgement task. Set in an alternate universe, all physical objects explode when they touch one another. Participants must read a description of an event and decide whether anyone or anything “touched” and, therefore, exploded.
Leah jumped the morp fracky? (Yes, this is an actual sentence in the study)
What is a morp fracky? The meaning of a sentence associated with a verb may depend on the words around it. “Leah rolled the ball,” implies that someone (Leah) applied force to the ball, causing it to move. However, “The ball rolled” does not. To avoid the effects of general world knowledge, we replaced the words surrounding the verb with nonsense words (e.g., Leah jumped the morp fracky).
So, what did we find?
Phase 1 of the project focused on verb classes comprised of 461 verbs and 7 groups of meaning. Through crowdsourcing, we collected 162,564 total judgements from 1,983 volunteers!
In order to ensure our tasks were reliable, each task was tested and redesigned until participants generally agreed with each other on what the “correct” answer was. Because large numbers of participants were recruited, this agreement was measured by calculating the most common response.
In terms of the actual results of the studies, consistency scores were lowest for the Evaluation task. Consistency was much higher in the other tasks, with the highest being the Physical Contact task. Researchers are still investigating whether the items that stray away from consistency percentage represent true differences in meaning, or potentially do not reflect anything at all.
The Future of VerbCorner
Results from Phase 1 provide support for an interaction between verb meaning and sentence type, but more work is needed in order to determine how strong this conclusion really is.
So, what’s next? Phase 2 of the project added over 10 new verb classes to our research! Phase 3, which also includes new classes and relations between sentence meanings, has just recently been launched!
If you’d like to learn more about VerbCorner, check out this blog post or our paper on The VerbCorner Project!
