Conventional Language Motivates Imitation, to an Extent

A desire to fit in with others is a seemingly natural part of being human. From a very young age, there is a strong motivation for children to follow conventions (e.g., Haun & Tomasello, 2011; Haun et al., 2012). Parents attend to conventional language and instruct their children to pay closer attention and to copy more precisely when conventional language is used (Clegg & Legare, 2017).

One interesting behavior that children engage in, which may result from this desire to adhere to convention, is overimitation. Horner and Whiten (2004) revealed that children would frequently slavishly copy all the actions that an adult model took to achieve a goal (such as retrieving a reward from a puzzle box), even if those actions were not causally necessary. Overimitation is common in children from cultures around the world (Nielsen & Tomaselli, 2010).

Research into why children overimitate suggests that they may do so because they think the model wants them to (e.g., they imitate to affiliate with the model) (Over and Carpenter, 2012). Or they may do so because they think it is the norm/ convention to do so (Clegg and Legare, 2016). Indeed, children overimitate more when conventional language (e.g., “Everyone always does it this way”) is used than when instrumental language (e.g., “I am going to…”) is used. Stengelin et al. (2022) found that overimitators are more likely to use normative language to teach a novel (non-instrumental) game to peers.

There are limits to children’s overimitation, however. For example, Keupp et al. (2016) found that children overimitate less when a causally irrelevant action is instrumentally focused or leads to the destruction of a valuable object. And children are also less likely to overimitate when they themselves act in a different context from the initial demonstration (Keupp et al., 2015). Similarly, Gergely and colleagues (2002) reported that children consider the context of the model’s actions before deciding whether or not to copy seemingly unusual actions.

Tools for cookie crushing: inefficient option on the right

Source: Cara DiYanni

What happens if a model performs a questionable action, such as intentionally selecting an inefficient tool to complete a task? In a previous study, we presented children with a model who purposely chose to crush a cookie with an object composed of fuzzy pom poms rather than selecting a more sturdy alternative (DiYanni et al., 2022) or chose to move rice with a flimsy rubber square as opposed to a sturdy rubber cup. Half the time, the model presented the task with conventional language and half the time with instrumental language. Results showed that the children imitated the model’s use of the inefficient tool significantly more often when conventional language was used than when instrumental language was used.

Tools for moving rice: inefficient option on the right

Source: Cara DiYanni

Our results also suggested that children in the conventional condition were more likely to teach a third party to use the inefficient tool. Did this mean that they were interpreting the inefficient tool as the one that was “for” the task? Ronfard and colleagues (2016) suggested that if children are asked to demonstrate the fastest way to perform a task, their subsequent actions would clarify how they interpret their observations of a model. We, therefore, wondered how children would respond in a pressure-based situation. For instance, if children were required to crush multiple cookies while we timed them for speed or to move rice from one bowl to a “fill line” on the other “as quickly as possible,” which tool would they choose?

Source: Cara DiYanni

Lyons and colleagues (2007) found that children persisted in overimitating unnecessary actions, despite an experimenter noting that they needed to achieve the task quickly. If children who initially copied the model’s use of an inefficient tool in our paradigm were to stick with that tool even in a pressured situation, it might indicate that they had truly encoded the inefficient tool as the one that is for the task. If, however, they switched to the more efficient tool in the pressured situation, it might indicate that an initial intention of trying to affiliate with the model or to conform to what the model presents as the convention is overridden when efficiency becomes more important in a timed situation. Additionally, if the model’s initial use of conventional vs. instrumental language impacted the children’s decisions in either the initial or the speeded trial—or both—it would tell us something about the power of language in children’s tool learning.

Source: Cara DiYanni

After each video, the child was asked to complete the task in an initial trial (“Which one would you need to crush the cookie/ move the rice?”). After the distractor task, the child was presented with the speeded trial. In each case, both tools (inefficient and efficient) were available for children to choose from. For the cookie-crushing task, the child was presented with four cookies and told, “You see, we have four cookies here. I’m going to time you; I want you to crush all of these cookies as quickly as you can. Are you ready? Go!” For the speeded trial of the rice-moving task, the child was presented with a bowl of rice and a second bowl that had a red line around the perimeter about 1/5 of the way up. They were told, “You see, we have some rice in this one bowl here. I’m going to time you; I want you to move rice from this bowl into the other bowl so that you fill this other bowl to the red line with rice as quickly as you can. Are you ready? Go!”

Figure 1: Percentage of Children Copying the Model's Selection of the Inefficient (NA) Tool in the Initial Trial

Source: Cara DiYanni

Source: Cara DiYanni

As can be seen in Figure 1, imitation rates in the initial trial were high. 67.4% of children imitated the model’s selection of the non-affordant (NA) flimsy rice tool, 74.4% of children copied the model’s choice of the NA, fuzzy cookie tool, 65.5% of children imitated the model’s use of the NA tool in both trials, and 77.9% of children imitated at least once.

Child Development Essential Reads

Why Are Food Allergies Becoming So Common?

How Early Do Infants Understand Humor?

Figure 2: Percentage of Children Copying the Model's Selection of the Inefficient (NA) Tool in the Speeded Trial

Source: Cara DiYanni

When it came to the speeded trial, however, rates of imitation dropped sharply, as seen in Figure 2. Only 20.9% of children used the NA rice mover, and only 30.2% used the NA cookie crusher. Just 25.6% of children imitated in one of the two trials, and a mere 11.6% of children imitated in both trials. Among the children who had imitated initially in the rice moving task, just 31% of them stuck with the NA tool in the speeded trial. This number was 40.6% for the cookie-crushing task.

These results to date suggest that while conventional language certainly has an initial impact on children’s decisions and influences a large number of children to copy a model’s selection of an inefficient tool, in a pressured time-based situation, children are more likely to switch to a tool that is more efficient for the task. One might argue that this is because they learned from the initial trial that the NA tool did not work, but our previous research, which included a teaching trial after the initial trial, suggests this is not the case. In our teaching trials, children who had heard conventional language tended to stick with the NA tool when instructing another how to do the task, even after potentially learning that it was not very effective when they themselves had tried it. However, the introduction of the pressured situation here led to a sharp drop in imitation rates.

In the coming months, we will collect data for the instrumental condition. It will be very telling to see whether rates of imitation are lower in the initial trial (as has been found in the past), but also to see what happens in the speeded trials. Will children’s rates of imitation look similar to those found here after hearing conventional language? Or might they drop even further? It is worth noting that between 1/5 and 1/3 of children still did use the NA tool in the speeded task after hearing a model use conventional language. If these numbers drop to near-zero after hearing instrumental language, it might indicate that although the effect of conventional language is diminished in a pressured situation, it does not disappear entirely.

Special thanks to Katie Haas, Nia Violette, and Ruth Gobena for their assistance with participant recruitment, data collection, and data entry and to all the children and parents who have participated in this study.