- 196 Downloads
Definition and Description
An experiment in which subject responses to purposeful (intentional) and unplanned (accidental) actions by an experimenter are measured as a means to identify whether an individual (human or nonhuman) understands the intentional actions of others. The intentional and accidental experiment is relevant to the topic of theory of mind, as it tests whether subjects understand intentional behaviors, one of the requisite components to understanding that others have thoughts and beliefs similar or separate to their own.
Two common variations of the accidental/intentional experiment use “accidental” and “intentional” actions of experimenters while indicating the location of rewards and “clumsy” and “competent” experimenters responsible for delivery of rewards. The basic premise of the experiment is that the subject must discriminate between the actions of the experimenter that will result in successful delivery of the reward and those that will not result in delivery of the reward.
In an initial study, subjects were trained to identify or request an item (usually food reward) hidden under an opaque container that has been identified with a marker placed on top of it, signaling the location of the food (Call and Tomasello 1998). During the subsequent testing phase, the marker was either intentionally placed on top of the container or accidentally dropped onto one of the other containers. In this procedure, the container was said to have been marked intentionally or accidentally. The subject was presumed to have recognized the intentions of others if the intentionally marked container was selected over the accidentally marked container (Call and Tomasello 1998).
In another variation of the task developed by Povinelli et al. (1998), during the training phase, subjects were exposed to a “clumsy” experimenter who accidentally dropped the reward, a “competent” experimenter who successfully delivered the reward, and a “negligent” (and in some cases “aggressive”) experimenter who intentionally dropped the reward. During the subsequent testing phase, the subject was to be given the chance to select the reward from one of the three experimenters. In this procedure, the subject was said to have recognized the intentions of others if the “competent” experimenter was selected over the “clumsy” and “negligent” experimenters (Povinelli et al. 1998).
In the intentional and accidental action experiment, the actions are often signaled by experimenters with eye gaze (toward or away from a location), intentional or accidental dropping or misplacing of the food reward, failure to deliver a food reward, and verbal cues such as “oops,” “oh no,” “okay,” and “there.”
Research on the ability of children to recognize intentions has primarily focused on young children, and the comparative framework for comparing humans and nonhumans involves studies with young, preverbal children (infant to 2 years). In these studies, researchers have relied on nonverbal measures such as eye gaze (Phillips et al. 1992), habituation-dishabituation (head turn) (Gergely et al. 1995), imitation (Meltzoff 1995; Carpenter et al. 1998; Olineck and Poulin-Dubois 2005), and reaching/grabbing (Behne et al. 2005) in order to identify whether or not children are able to recognize the intentions of others. Research using each of these measures suggests that by the end of their first year, children are able to identify the intentions of others.
Many of the studies that examine the ability of children to recognize intention and accidental actions incorporate verbal indicators or cues to signal the intentional and accidental actions. For example, Carpenter et al. (1998; replicated by Olineck and Poulin-Dubois 2005) examined the ability of young children (14–18 months) to reproduce or imitate intentional and accidental actions modeled by an adult. These actions were indicated by the experimenter as intentional or accidental with the words “there!” (intentional action) and “woops!” (accidental action). Research using verbal cues to signal intentional and accidental actions indicates that young children, 16–18 months of age, imitated significantly more intentional than accidental actions modeled by an adult. Additionally, in studies that require the child to imitate an adult model as a measure of understanding, the intentional and accidental actions were signaled by the modeler/experimenter by failing to complete the task, negative facial expression (frown), dropping the object, or refusing to offer the object to the child (Meltzoff 1995; Behne et al. 2005). Results of these studies indicate that children as young as 9 months of age are able to differentiate the intention of others by observing the modeler/experimenter’s actions and facial expression (Behne et al. 2005).
The bulk of research employing the use of intentional and accidental experiment methodology to test intentionality is found in research with primates, specifically the great apes. Two examples from research on chimpanzees and orangutans are presented below.
Call and Tomasello (1998) investigated the ability of human children, chimpanzees (Pan troglodytes), and orangutans (Pongo pygmaeus) to identify the intentions of others. The procedures for eliciting responses to intentional and accidental behaviors of the models consisted of a training and a testing phase. In the training phase, the authors trained the subjects to select a “correct” box containing a food reward using a three-phase delayed stimulus discrimination task. Phase one consisted of introducing the subjects to the concept that a food reward could be found under one of the three boxes. Training subjects included partially hiding the food reward under the boxes. Visibility of the food rewards was slowly decreased on each subsequent trial until fully hidden. Phase two consisted of the introduction of a marker that “marked” the box containing the food (stimulus discrimination phase). In this phase, the subject was trained to select the box with the marker in order to obtain the hidden food reward within. The subjects learned to associate the marked box with the location of the food reward. Phase three consisted of a delayed-response procedure. In this phase, the marker was placed on the box containing the food reward but was then removed. The subject was then allowed to select the box with the food reward after a delay. Responses of the subject were delayed 1 s, increasing in successive trials to 10 s and then to 20 s (delayed discrimination task).
During the testing phase, the experimenter baited the boxes out of view of the subjects and then allowed the subjects to view the boxes without the presence of the markers. Two intentional actions and two accidental actions were then performed. The order of the presentation of the intentional and accidental actions, as well as the location of the hidden food, was counterbalanced. The two intentional actions consisted of (1) the experimenter picking up the marker and placing it on the top of the box while looking at the box and (2) the experimenter picking up the marker and dropping it on the box while looking at the box. In both intentional action types, the marker was left on the box for 3 s and was then removed. The boxes were then pushed toward the subject to allow the subject to select the box containing the food reward.
The two accidental actions consisted of placing the marker on a ledge behind the boxes and (1) the experimenter moving the ledge resulting in the marker being dislodged and falling on to one of the boxes and (2) the experimenter knocking the marker off of the ledge on to the box. While performing the accidental actions, the experimenter did not look at the boxes. After the accidental placement of the marker, the experimenter left the marker in place for 3 s, then removed the marker, and allowed the subjects to select the box containing the food reward. A verbal prompt “oops” was used for two of the chimpanzee subjects in order to investigate the potential benefit of a verbal prompt on performance. Each trial consisted of one intentional action and one accidental action. The intentional and accidental placement of the marker may or may not have corresponded to the correct location of the reward. In other words, the marker may or may not have signaled the actual location of the reward. The authors hypothesized that when presented with the choice of these two marked locations, the intentionally marked location should be selected over the unintentionally or accidentally marked location, indicating the individual understands the intention of the experimenter. The subjects (both chimpanzees and orangutans) tended to select the intentionally marked boxes more than the accidentally marked boxes at the beginning of the experiment, and subsequently accuracy of their selections decreased as the experiment progressed. With regard to the decreased accuracy, the authors concluded that “…[this] could be interpreted as an indication that accidental actions became less plausible after accidents occurred in each trial” (Call and Tomasello 1998, p. 202).
Povinelli et al. (1998) used a procedure in which young chimpanzees were trained to point to unfamiliar actors in order to receive rewards (cups of juice). During the training phase of this experiment, the subjects were exposed to actors who handed cups of juice to experimenters who then offered the juice to the chimpanzees. The chimpanzees were then presented with actors who (1) “accidentally” dropped the juice on the floor before handing it to the experimenter, (2) “intentionally” poured the juice on the floor, or (3) “aggressively” threw the juice on the floor. The chimpanzees were then tested to see which actor they selected for delivery of the juice as indicated by pointing to the actors. The authors hypothesized that the actors who accidentally dropped the juice on the floor should be selected over those who intentionally or aggressively dropped the juice. The results showed nonsignificant effects of the three conditions, indicating that the chimpanzees did not show a preference for which actor delivered the juice, leading the authors to conclude that chimpanzees are not able to tell the difference between intentional and accidental actions of human experimenters (Povinelli et al. 1998, p. 215).
In a second experiment, Povinelli et al. (1998) addressed some methodological issues such as the possibility that the chimpanzees preferred interacting with certain human actors over others. In this experiment, the actors were counterbalanced across three conditions for each chimpanzee; the same actors played one of the three roles for the same chimpanzee but switched roles for different chimpanzees. In this experiment, a neutral actor, victim actor, and intentional actor were used to deliver (or not deliver) a food reward. Each session contained two combinations of actor scenarios. The neutral actor picked up the food reward and handed it to the experimenter who then handed it to the chimpanzee. The victim actor delivered the food reward to the experimenter who then ate the food reward intended for the chimpanzee. The intentional actor always ate the food reward intended for the chimpanzee rather than handing it to the experimenter. After a training session in which the chimpanzees were exposed to each actor multiple times, counterbalancing for action and location (left or right side), the chimpanzees were tested for preference of human actors by exposing the chimpanzees to combinations of actors as (1) intentional actor versus victim, (2) victim versus neutral actor, and (3) intentional actor versus neutral actor. Results indicated that the chimpanzees did not prefer one actor over another, regardless of their behavior toward the experimenter. Similar to the first experiment by Povinelli et al. (1998), results from this modified procedure indicated that chimpanzees do not recognize the intentions of human experimenters.
As suggested by Call and Tomasello (1998), the accidental and intentional action studies used with nonhuman subjects may allow us to investigate whether nonhumans are able to differentiate between the subtle behavioral (and in some cases verbal) cues that accompany intentional and accidental actions made by humans. However, just as errors occur in the ability to recognize intentional or accidental acts in human-to-human interaction (e.g., sarcasm, deception, threats, displays of emotion, etc.), we would also expect similar errors in nonhuman subjects, and, similarly, failure should not necessarily be an indication of inability.
Other limitations suggested by Povinelli et al. (1998) point to the developmental stage of the subjects tested. In their study, the chimpanzees (5–7 years old) were unable to distinguish between the accidental and intentional acts presented by the unfamiliar actors. Povinelli et al. suggest that this may have been due to the lack of cognitive development of their subjects. The authors also suggest that their subjects may not have perceived the difference between the types of actions “…because the severity of the outcome was too mild” (p. 215). Povinelli et al. seem to suggest that the consequences of the actor’s actions were of little significance to the subjects possibly due to the low or diminished value of the food reward.
Perhaps the most obvious limitation of the accidental and intentional action experiment is the use of linguistic cues employed by many researchers. Exclamations of “whoops” to indicate accidental action or “okay” to indicate intentional action may carry little to no meaning for the nonhuman subjects in these experiments. Call and Tomasello (1998) found that the two nonhuman subjects that were exposed to verbal cues did not performed any better than did subjects that were not exposed to verbal cues. Whether or not an individual recognizes an exclamation of “whoops” or “okay” to indicate accident or intention, the linguistic cue may simply serve to draw the attention of the subject to a desired reward, resulting in negative results. For example, inadvertently dropping a delicious treat on the floor may, for many test subjects in a laboratory environment, signal a rare opportunity for acquiring the delicious treat regardless of the intention of the actor. The verbal cues used to indicate accident or intention simply indicate the availability of the food reward.
Consideration must be given to the early rearing environment and the role this may play in the ability to understand the accidental or intentional acts of others. Call and Tomasello found that the orangutan Chantek performed much better than the other subjects in their study of accidental and intentional actions. Chantek’s early rearing experience resembled that of a young child from the age of 9 months to 9 years. Chantek acquired the signs of American Sign Language and spent most of his waking hours with human caregivers (Miles 1986). Call and Tomasello attributed Chantek’s superior performance to these early rearing experiences and human enculturation. In addition to the individual’s environment, knowledge of the materials used in these tests as well as the familiarity of the animals to the experimenter/caregivers may yield contradictory results (although see Pitman and Shumaker 2009, refuting differences in early rearing conditions).
The social behavior of the different primate species tested must also be taken into account when assessing the ability to recognize the intentions of others. For example animals in the wild likely reap the rewards brought about by the actions of others, regardless of the intention of the act. The inadvertent (or possibly intentional) dropping of a highly coveted food item by a dominant individual may be recovered by a subordinate individual. Knowledge of whether the dominant individual meant to drop the food item may be irrelevant, if even knowable; the opportunity to recover the food and avoidance of repercussion from the dominant individual are paramount and perhaps necessary for survival. Another example of the behavioral differences between human and nonhuman primates was suggested by Call and Tomasello (1998). The authors stated that the type of action that resulted in the subjects selecting the location of the hidden food reward was the dropping of the marker on the box and not the intentional placement of the marker on the box. The authors hypothesized that the subjects identified this action (dropping) as an indicator of the hidden food reward because this pattern of behavior is more common (dropping of food) in chimpanzees and orangutans than actual sharing of food by handing food or placing of food in another’s hand.
Although the results of the accidental and intentional experiment may yield contrasting results across, and in some cases within, human and nonhuman populations with regard to evidence of theory of mind, the experimental procedure provides some insight into the cognitive capacities of these species. Age, social behavior, and methodological considerations however must be considered when comparing populations and making conclusions regarding the capacity to identify the intentional and accidental actions of others.
- Miles, H. L. (1986). How can I tell a lie? Apes, language and the problem of deception. In R. W. Mitchell & N. S. Thompson (Eds.), Deception: Perspectives on human and nonhuman deceit (pp. 245–266). Albany: State University of New York Press.Google Scholar