Micah B. Goldwater

Causal Systems Categories: Differences in Novice and Expert Categorization of Causal Phenomena

We investigated the understanding of causal systems categories--categories defined by common causal structure rather than by common domain content--among college students. We asked students who were either novices or experts in the physical sciences to sort descriptions of real-world phenomena that varied in their causal structure (e.g., negative feedback vs. causal chain) and in their content domain (e.g., economics vs. biology). Our hypothesis was that there would be a shift from domain-based sorting to causal sorting with increasing expertise in the relevant domains. This prediction was borne out: the novice groups sorted primarily by domain and the expert group sorted by causal category. These results suggest that science training facilitates insight about causal structures.

Structural Priming as Structure-Mapping: Children Use Analogies From Previous Utterances to Guide Sentence Production

What mechanisms underlie childrens language production? Structural priming--the repetition of sentence structure across utterances--is an important measure of the developing production system. We propose its mechanism in children is the same as may underlie analogical reasoning: structure-mapping. Under this view, structural priming is the result of making an analogy between utterances, such that children map semantic and syntactic structure from previous to future utterances. Because the ability to map relationally complex structures develops with age, younger children are less successful than older children at mapping both semantic and syntactic relations. Consistent with this account, 4-year-old children showed priming only of semantic relations when surface similarity across utterances was limited, whereas 5-year-olds showed priming of both semantic and syntactic structure regardless of shared surface similarity. The priming of semantic structure without syntactic structure is uniquely predicted by the structure-mapping account because others have interpreted structural priming as a reflection of developing syntactic knowledge.

On the acquisition of abstract knowledge: Structural alignment and explication in learning causal system categories

This research studies a relatively unexplored aspect of expertise - the ability to detect causal relational patterns in multiple contexts - and demonstrates learning processes that foster this ability. Using the Ambiguous Sorting Task (AST), in which domain information competes with causal patterns, we previously found that science experts spontaneously noticed and sorted by causal patterns such as positive feedback, while novices sorted primarily by content domain. We investigated two kinds of learning experiences that we claim are needed to achieve high fluency in detecting key cross-domain patterns. We found that direct explication of example phenomena increased peoples accuracy in depicting the examples, but did not increase sensitivity to the causal patterns in new examples. However, analogical comparison between parallel examples did lead to greater propensity to detect the causal patterns across diverse examples. Combining within-example explication with between-example alignment led to the greatest gains in generalized sensitivity to causal patterns.

Categorizing entities by common role.

Many categories group together entities that play a common role across situations. For example, guest and host refer to complementary roles in visiting situations and, thus, are role-governed categories (A. B. Markman & Stilwell, Journal of Experiment & Theoretical Artificial Intelligence, 13, 329-358, 2001). However, categorizing an entity by role is one of many possible classification strategies. This article examines factors that promote role-governed categorization over thematic-relation-based categorization (Lin & Murphy, Journal of Experimental Psychology: General, 130, 3-28, 2001). In Experiments 1a and 1b, we demonstrate that the use of novel category labels facilitates role-governed categorization. In Experiments 2a and 2b, we demonstrate that analogical comparison facilitates role-governed categorization. In Experiments 1b and 2b, we show that these facilitatory factors induce a general sensitivity to role information, as opposed to only promoting role-governed categorization on an item-by-item basis.

What is typical about the typicality effect in category-based induction?

Research on category-based induction has documented a consistent typicality effect: Typical exemplars promote stronger inferences about their broader category than atypical exemplars. This work has been largely confined to categories whose central tendencies are also the most typical members of the category. Does the typicality effect apply to the broad set of categories for which the ideal category member is considered most typical? In experiments with natural and artificial categories, typicality and induction-strength ratings were obtained for ideal and central-tendency exemplars. Induction strength was greatest for the central-tendency exemplars, regardless of whether the central tendency or the ideal was rated more typical. These results suggest that the so-called “typicality” effect is a special case of a more universal central-tendency effect in category-based induction.

Constructional sources of implicit agents in sentence comprehension

Abstract Much research about on-line sentence comprehension focuses on the contributions of individual lexical items, with specific interest in verbs. One aspect of sentence meaning that has been claimed to be rooted in verb representation is event structure. There is a growing body of evidence supporting the claim that the verb is not the sole contributor of event structure, but that the syntactic construction of a sentence is also a contributor. In this paper, we repeat a study designed to support a verb-based view using novel verbs derived from nouns. The pattern of sentence comprehension is the same for both known verbs and novel verbs, suggesting that the syntactic construction of the sentence also contributes to event structure.

From Concrete Examples to Abstract Relations: The Rostrolateral Prefrontal Cortex Integrates Novel Examples into Relational Categories

Abstract The ability to form relational categories for objects that share few features in common is a hallmark of human cognition. For example, anything that can play a preventative role, from a boulder to poverty, can be a “barrier.” However, neurobiological research has focused solely on how people acquire categories defined by features. The present functional magnetic resonance imaging study examines how relational and feature‐based category learning compare in well‐matched learning tasks. Using a computational model‐based approach, we observed a cluster in left rostrolateral prefrontal cortex (rlPFC) that tracked quantitative predictions for the representational distance between test and training examples during relational categorization. Contrastingly, medial and dorsal PFC exhibited graded activation that tracked decision evidence during both feature‐based and relational categorization. The results suggest that rlPFC computes an alignment signal that is critical for integrating novel examples during relational categorization whereas other PFC regions support more general decision functions.

Learning of role-governed and thematic categories.

Natural categories are often based on intrinsic characteristics, such as shared features, but they can also be based on extrinsic relationships to items outside the categories. Examples of relational categories include items that share a thematic relation or items that share a common role. Five experiments used an artificial category learning paradigm to investigate whether people can learn role-governed and thematic categories without explicit instruction or linguistic support. Participants viewed film clips in which objects were engaged in similar actions and then were asked to group together objects that they believed were in the same category. Experiments 1 and 2 demonstrated that while people spontaneously grouped items using both role-governed and thematic relations, when forced to choose between the two, most preferred role-governed categories. In Experiment 3, category labels increased this preference. Experiment 4 found that people failed to group items based on more abstract role relations when the specific relations differed (e.g., objects that prevented different actions). However, Experiment 5 showed that people could identify them with the aid of comparison. We concluded that people can form role-governed categories even with minimal perceptual and linguistic cues.

Relational Discovery in Category Learning

Learning categories defined by the relations among objects supports the transfer of knowledge from initial learning contexts to novel contexts that share few surface similarities. Often relational categories have correlated (but nonessential) surface features, which can be a distraction from discovering the category-defining relations, preventing knowledge transfer. This is one explanation for “the inert knowledge problem” in education wherein many students fail to spontaneously apply their learning outside the classroom. Here we present a series of experiments using artificial categories that correlate surface features and relational patterns during learning. Our goal was to determine what task parameters and individual differences in learners shift focus to the relational aspect of the category and foster transfer to novel disparate exemplars. We consistently showed that the effectiveness of task structure manipulations (e.g., the sequence of learning exemplars) depended on the learners’ strategies (e.g., whether learners are oriented toward discovering rules or focusing on exemplars). Further, we found support that “inference-learning,” wherein learners are presented with incomplete exemplars and learn how to infer the missing pieces, is an effective way to promote relational discovery and transfer, even for learners who are not predisposed to make such discoveries.

Licensing Novel Role-Governed Categories: An ERP Analysis

Markman and Stilwell (2001) argued that many natural categories name roles in relational systems, and so they are role-governed categories. This view predicts instantiating a novel relational structure licenses the creation of novel role-governed categories. This paper supports this claim and helps to specify the mechanisms underlying this licensing. Event-related potentials were recorded while participants read passages of text. Participants instantiated novel relational representations by interpreting novel verbs derived from nouns during reading. Sentences later, comprehension of novel role terms derived from the novel verb was facilitated relative to a control condition where the novel verb was paraphrased using the root noun in its familiar form. This comprehension facilitation was marked by a reduced negativity elicited from the role term in the Novel Verb condition relative to the Paraphrase from 400 to 500 ms post-stimulus-onset. This relative difference in negativity is consistent with both the N400, which is a marker of semantic integration, and the Nref effect, which reflects the working memory load required to resolve reference. Additionally, because this increased negativity persisted until 670 ms post-stimulus-onset, and not that the Paraphrase condition elicited an increased positivity (i.e., the P600), we ruled out that the licensing effect is rooted in morphosyntactic processes.