 |
Roth, E.M., and Shoben, E.J. (1983). The effect of context on the structure of categories. Cognitive Psychology, 15, 346 – 378. Barsalou, L. W. (1983). Ad hoc categories. Memory and Cognition, 11 (3), 211 – 217 Ahn, W. & Bailenson, J. (1996). Causal attribution as a search for underlying mechanisms: An explanation of the conjunction fallacy and the discounting principle. Cognitive Psychology, 31, 82-123.
|
Editorial Paper: Context and Theory
Summary/Common Threads
Roth and Shoben were interested in how typicality ratings are affected by situations within specific contexts and in context-free situations. Participants were presented with anaphoric expressions and their reading time was measured. The results found that that reading times were fastest when the context sentence suggested the exemplar in the target sentence and slowest when the context sentence contained category information but the target sentence contained an unlikely exemplar. Thus, context was found to affect participants’ comprehension. The second experiment found that within a specific context, participants rate exemplars of categories using a goodness-of-example (GOE) distribution. The third experiment found that both concepts in isolation and in context have GOE distributions. However, the order of representativeness changes for the concepts within a context. Context influences which concepts will be rated as the best exemplars for a specific category.
Barsalou’s paper focused on ad hoc categories and whether or not they possess graded structure. Participants were given a background story containing a specific context and an ad hoc category label to match. They were then given six items and asked to rank them from most to least appropriate for that specific category. The results demonstrated that ad hoc categories possess graded structure. In a second experiment, participants were given both ad hoc and common categories and asked to list exemplars for each. More variability was found in the exemplars listed for the ad hoc categories than in the common categories, but graded structure was apparent for both. Lastly, participants were given exemplars for common and ad hoc categories and asked to provide typicality ratings. Typicality ratings were very similar for both types of categories.
Ahn’s paper addressed the conjunction fallacy and the discounting principle, which can both be used to explain the causes of events. The results of the first experiment found support for the mechanism approach, which posits that people construct mechanism-based explanations rather than covariation explanations when explaining events. The conjunction effect was stronger with mechanism-based explanations than with covariation-based explanations. The discounting effect was also stronger with mechanism-based explanations. In the second and third experiment, participants were presented with pairs of explanations and context cues, which served to link the explanations together or to bias the participants against one of the explanations. Participants performed either a conjunction or a discounting task. The results were again supportive of the mechanism approach. Based on these results, it seems that when people make judgments about causal relationships, information about the underlying mechanisms is more important than covariation information.
These three studies demonstrated the dramatic effect context has on various aspects of cognition. They were grounded in different bodies of research, specifically the formation of categories, the explanation of events, and the processes involved in reading, yet in all three, context was found to play a very influential role. Changing the context in which a concept is housed alters its interpretation and categorization. People look to context to help them process information and make judgments. Even when people are not consciously aware of context, it still affects their categorization and concept formation.
Essentially, context represents a form of an ad hoc category which people use to evaluate and understand new information. The context in which a person learns a concept serves as its category until the concept is integrated into another category. The Roth & Shoben and Barsalou articles both provide interesting perspectives on common versus ad hoc categories. It is taken for granted that people understand these types categories since they use them often, yet when asked to think about the category structures, people realize that they are very simple. A person may believe that she has a clear-cut category for a group of concepts, but as these studies demonstrate, category boundaries shift when context changes, making judgment about category membership more complicated. The Ahn article did not directly address categories, however ad hoc categories can be applied to their findings. When people are evaluating causal relationships, they are basically evaluating events that believe belong in a certain category. For example, people may have an ad hoc category for the causes of a car accident. Within this category concepts like bad weather, poor vision, and bad driving are housed. In order to evaluate the cause of the car accident, a person must activate her ad hoc category and compare her concepts to the available information. Depending on the context and situation, different forms of the ad hoc category will be activated and a person will make different judgments.
Class Discussion Points/Issues
The class discussed interesting questions and issues regarding the Roth and Shoben article. First, the use of an eye tracker in this study was addressed. Would including eye-tracking data have been important in this particular study? Would the results have been any different? For example, a word like bugs has the exact same spelling and part of speech, yet the meanings are completely different. Next, the class pondered if there are any situations in which both meanings of a word would be activated from the same sentence, possibly in an anaphoric sentence. Next, the study’s inclusion of a “no context” condition was also debated. The question of whether or not people can generate exemplars and create categories in the absence of context was discussed. One argument is that context is always present. For example in the sentence, “I saw a bird”, although there is no specific context included in the sentence, there is a broad overarching context, which may be activated. This permanent, long-term context may influence a person’s conceptualizations and categorizations and thus affect her reading and comprehension. It is possible that the “no context” condition was really a condition without a specific context but still contained a general context which people used in their judgments.
The discussion for the Barsalou article focused around he basic question of whether or not graded typicality really occurs in ad hoc categories the same as it does in common categories. Ad hoc categories are different than common categories in many ways including their creation and use. The results support the existence of graded structure for ad hoc categories, but many questions still arise. Are there are certain situations in which ad hoc categories do not have a graded structure? For example, in the ad hoc category of ingredients needed for baking a cake, the ingredients are all necessary, none are more important than others, so, can you put them into a graded structure? However, the ad hoc category of things you need to pack for your vacation fits easily into a graded structure with some items being important and/or necessary than others. The question of hierarchy in ad hoc categories was also discussed. The ad hoc category of a “to do list” can be both hierarchal and non-hierarchal depending on the situation. In both cases, the category’s members are things that need to be completed, but some lists contain items that are given priority over other items, while other lists contain items, which are equally important and must be completed during the same time span. Lastly, the goal-oriented nature of ad hoc categories was discussed (eg. things you need for an overnight guest). Do all ad hoc categories have this nature? People use both ad hoc and common categories very often in their daily lives.
The Ahn article provided a different kind of discussion, as it was somewhat distinctive among the other articles in that its relationship to concepts and categories was not as clearly defined. The class discussed the advantages of the two types of information discussed in the article: mechanism based and covariation based. Mechanism based explanations can be very satisfying for most people and when given this information they will not seek further understanding. Covariation based explanations, however typically require additional information be gathered before people feel satisfied. The class spent a lot of time discussing correlations and the situations in which a correlation is an acceptable explanation for an event. One such example is a myth. Myths were created to help people understand why certain events occurred together. Without any scientific support, myths become ingrained in the culture and are readily accepted as suitable explanations. Lastly, the class discussed human’s driving motivation to understand the cause of an event or situation. This knowledge, even if it is fallacious, provides us with a sense of security and control. For example, most individuals do not know how their computer works, but when it freezes, they are motivated to understand the cause. Any correlation they can draw to help them is typically accepted as a plausible explanation and they will use this information in future to avoid the problem, providing them with a sense of control over the computer.
Directions for Future Research
In all of these studies, context was provided to participants in a written format. It would be interesting to present contexts in different forms and to examine the pattern of results. For example, participants could be shown a picture or hear a story as the way of presenting context. Another interesting manipulation would be to examine the effects of context using a foreign or made-up language. Using the Roth et al. study paradigm, participants could be taught a made-up language and after demonstrating mastery, the same experiment could be conducted. The effect would probably be diluted if it was found at all, but it would be an interesting manipulation.
Examining the effects of context in different, contrasting cultures is another direction for future research. One would predict that context would not have as strong of an influence on members of a culture who are not familiar with the presented context. For example, the Barsalou paper uses the ad hoc category of “ways to escape being killed by the mafia”. This ad hoc category is situated in a very specific American context. Other cultures would probably have a much more difficult time generating exemplars and providing typicality ratings for this ad hoc category. The role of familiarity itself as it relates to context should also be examined in future work.