Studies motivated by cognitive load theory experiments (e.g., Ward & Sweller, 1990) suggest that problem solving can interfere with learning, and show that alternative exercises such as studying worked examples often lead to better performance. However, problem solving at has often been found to have benefits for both retention and transfer (e.g., Jacoby, 1978) and some studies (e.g., Charney & Reder, 1986) show better learning from problem solving than from worked examples. Many factors may affect the outcome of such studies. Often the methods that produce difficulties during training seem to lead to the best long term performance (Schmidt & Bjork, 1992) Some evidence suggests that the least capable learners benefit most from relatively active, unstructured training methods (e.g., Hermann, 1969; Whitener, 1989) but others hold the opposite opinion (e.g., Snow, 1989). People often fail to make use of examples unless stimulated, for example, by provision of multiple, varied examples (e.g., Gick & Holyoak, 1983). An experiment comparing three approaches to instruction was conducted in an effort to clarify the circumstances favoring problem solving or study of examples in training. Subjects either first attempted to solve a problem then studied a related example, studied an example then attempted to solve a related problem, or studied a pair of related examples. Problem solving first resulted in slower and less accurate performance during training, as did studying less similar pairs of exercises. Subjects who studied worked examples without problems performed poorly on an immediate test, and the least capable among them had particular difficulty with the most complex tasks. On a delayed test, among subjects who studied the less similar pairs of exercises, those trained with pairs of worked examples performed more poorly on difficult problem components than did problem-first subjects. On the delayed test, among trainees of average or lower intelligence, those trained with problems first were quickest. Consistent with the ideas of Schmidt and Bjork (1992), conditions that produced difficulties during training produced the best ultimate performance. Contrary to Snow's (1989) conclusion, this seemed to be particularly important for the least capable trainees.

Two experiments investigated the effect of layout complexity for performance at varying levels of practice on four types of information extraction tasks. Layout complexity is defined as the number of unique horizontal and vertical starting positions of items in the display (Tullis, 1984). Tullis investigated the ability of display-formatting variables (overall density, local density, number of groups, size of groups, number of items, and layout complexity) to predict human performance and preference. Although layout complexity was the best predictor of subjective ratings, it did not contribute to the prediction of search time beyond what could be predicted by overall density, local density, number of groups, and size of groups. This is a particularly interesting finding since vertically aligning lists and left-justifying items, practices strongly recommended in guidelines for display formatting (e.g., Engel & Granada, 1976; Smith & Mosier, 1986), are important factors in Tullis's definition of layout complexity. Thus, the guidelines and Tullis's model lead to conflicting predictions concerning the effect of left-justifying text on user search time. In the first study, layout complexity was manipulated by either left-justifying or not left-justifying text. Although the text was not left-justified, the starting positions of the text were ordered rather than random. In the second study, subjects viewed a third experimental screen that displayed the starting positions of items in a completely unpredictable pattern. Subjects performed all four tasks (find label, scan data, compare labels, and compare data) in four one-hour sessions. Moderate violations of the typical guideline recommendations did not increase user search time across all four tasks in either the first or the second study. However, when subjects compared multiple data values, the random format did increase user search time. Though performance using the three experimental screens was comparable across the four tasks with only one exception, subjective ratings demonstrated differences between the three formats. Subjects disliked the random format and degraded their performance using the random screens.

Transfer of training is the degree to which a skill or principle learned in one environment can be applied in another environment. Most research that demonstrates transfer relies on the use of hints or explicit instructions identifying the applicability of the information learned in training to the test task. Critics charge that this is not really transfer at all, but simply following instructions. The research reported herein describes an efficient means for testing hypotheses in a fault diagnosis task that, although it would appear to be an obvious strategy, requires an extremely simple training task for subjects to detect. Subjects in Experiment 1 apply the learned principle to a slightly more complex but similar problem, demonstrating near transfer. Subjects in Experiment 2 apply the principle in a completely dissimilar task, exhibiting far transfer.

Virtual Reality (VR) has been deemed the interface of the future. However many people get lost navigating through a simple DOS hierarchy due to low spatial ability. Research has shown that spatial ability can be increased through training, and that there is a link between spatial ability and VR. This experiment attempts to replicate and combine those studies by trying to improve the subjects' spatial ability and increase the efficiency of a VR interface. Thirty subjects were pre- and posttested on their spatial ability, as well as their ability to navigate through a virtual environment. Half the subjects underwent 15 hours of training using seven spatial ability tests. The trained subjects improved their spatial ability, not just on those seven tests, but in general. There was no significant correlation between spatial ability and ability to navigate in VR, and no difference in improvement in that ability between the two groups.

Researchers often assume that automatic semantic priming effects found in lexical decision tasks are due to spreading activation at a conceptual knowledge level. Recent research questions this assumption by demonstrating that automatic priming does not occur for word pairs that are semantically related but not normatively associated (e.g., Shelton & Martin, 1992). The present research investigated priming at a conceptual level by manipulating depth of prime processing which forced subjects to access semantic knowledge. To ensure the priming effects were due to automatic processing two tasks were used which were designed to lessen the likelihood that subjects would engage in strategies making use of the semantic relationship between the word pairs. The first task used a single word presentation procedure and a low proportion of related words in the stimulus set which has been shown to uncover automatic priming in a lexical decision task (Shelton & Martin, 1992). Several experiments were conducted where subjects were required to make semantic decisions about each of a sequence of words. The second task was a modified Stroop task in which subjects either processed the prime nonsemantically or semantically and named the color of a subsequent target word. Previous studies have shown that an associative relationship between the prime and target slows naming of the color of the target. It has been argued that the inhibition in this task results from automatic processing because noticing prime-target relationships would only produce greater interference. The results showed that priming was found for semantically related, unassociated word pairs in the semantic judgment tasks but not in the modified Stroop task. A final experiment demonstrated that the relatedness priming found in the semantic judgment experiments only came about when the prime-target pairs were related and the decision between the prime and target was the same. The results from the semantic judgment experiments suggest that the priming effect is due to the similarity in the information used to make the decision rather than activation of all semantic features shared between two concepts. The results from the modified Stroop task suggest there is no immediate influence of conceptual information on lexical access.

Consistent biases in the perception of risk have been demonstrated by research. An imaging task was used in this study in an attempt to change those perceptions. The task consisted of sketching scenarios depicting an accident involving eight different hazards. The task effected a change in self risk ratings taken before and after the task, but the change was different for each of two groups of hazards. The hazards were divided into a high and a low group based on the pre-task self risk ratings. Lawnmower, toaster, coffee maker, and bicycle were found to be initially perceived as low risk. Pesticide, swimming pool, asbestos, and power saw were in the initially perceived high risk group. Pre-post self risk rating change scores showed an increase for the low risk group and a decrease for the high risk group. The interaction between sketching and not sketching an accident and the initial perceptions of low and high risk was significant. This finding lends hope that the accident scenario imaging method can be used to correct biases in the perception of risk of certain hazards. An additional analysis addressed the relationship between the perception of risk and other cognitions. Self risk ratings were found to be more closely related to probability estimates for injuries to self than to injury frequency estimates for the general public. Frequency and probability estimates do not tell the whole story, however, as far as risk perception is concerned. Perceptions for the severity of the consequences of an injury caused by the hazard, the catastrophic potential of the hazard, the ease of imagining an injury caused by the hazard, and the extent of exposure to the hazard are also significantly related to the perception of risk to self presented by the hazard. Important differences were observed in this study between perception of risk for self and for others. These findings highlight important considerations in the study and measurement of risk perception which is an important part of any effort to de-bias risk perception. When risk perception is found to be biased in a potentially harmful manner, an understanding of the role of the cognitive availability of accident scenarios and how to change that availability might then be used to correct those biases.

Many researchers have assumed that testing for semantic priming in a lexical decision task assesses whether aphasic patients have preserved automatic semantic processing. However, research with normal subjects has shown that priming in lexical decision may result from controlled processing (e.g. Balota & Lorch, 1986). The present research developed a lexical decision task that would reflect automatic priming and tested aphasic patients on this task. Also, the nature of the semantic knowledge revealed through priming was investigated by examining priming for words that are semantically related but share no associates (e.g. poem-novel). Contrary to the findings of Milberg & Blumstein (1981), automatic priming was obtained for a Broca's aphasic. Also, priming was not revealed for semantically related words sharing no association, indicating that automatic priming relies on lexical associations between words, not semantic knowledge. Automatic semantic priming does not appear to be a useful measure for assessing semantic knowledge organization.

Most theories of language recognize the importance and complexity of verbs. Verbs convey both semantic and syntactic information, and therefore, the study of verb representation and processing can provide important insights into the nature and relationship of semantic and syntactic representation within the language processing system. Two models of verb representation were examined: (1) Semantic Complexity (Johnson-Laird, 1983); and (2) Conceptual Constituents (Pinker, 1989). The validity of these models was investigated by examining patterns of verb disruption in seven language impaired individuals. Investigation of patients' verb processing did not provide evidence for a model of verb representation based on semantic complexity, however, it did provide support for semantic representation of thematic roles within the lexicon separate from syntactic representation as posited by the conceptual constituent model. These findings were followed by an experiment investigating whether the language processing system has a bias for particular thematic roles in specific syntactic positions. The results indicated no bias for the syntactic position of specific thematic roles.

Previous research on the human process of debugging indicates that it can be described as a process of testing hypotheses. It is noted that Hypothesis Testing Theory (HTT) as proposed by Levine (1966) can provide a guiding framework to study debugging. Experiments 1 and 2 tested the following hypothesis: Given a very simple debugging task for which there is a clearly defined optimal strategy for solution, would people discover and use that strategy. Subjects tended to discover and use the optimal strategy in experiment 1, but not for the slightly more complex task of experiment 2. Experiments 3 and 4 tested the assumptions of HTT of domain sampling on a task similar to the ones used in Experiments 1 and 2. The assumptions of HTT were supported by the results of experiment 3. All of the subjects in experiment 4 used a sequential search strategy. The results of each experiment are discussed and additional research is proposed.

Although use of the mental model construct has proliferated in recent applied research, the construct lacks an agreed upon method of measurement. Importantly, the validities of the different measurement techniques have not been established. The purpose of this research was to evaluate the validity of several mental model measurement methods, with the criterion being task performance. Additionally, mental models were examined in the context of a real-world, complex problem-solving situation. Four mental model measurement techniques were evaluated: a laddering structured interview, concept relatedness ratings, a diagramming structured interview, and think aloud while troubleshooting. Nineteen U.S. Air Force technicians varying in troubleshooting expertise each completed the mental model measures. In addition, the technicians each worked to troubleshoot a moderately difficult problem. The results indicate that two of the evaluated techniques were each independently predictive of troubleshooting performance: the laddering structured interview and the concept relatedness ratings. Because these measures are predictive of performance, it is recommended that researchers consider using these techniques in work requiring the measurement of mental models. The next step in this line of research involves a characterization of the kinds of information offered by different mental model measures. This research also revealed basic psychological issues regarding the development of expertise. Specifically, the results suggest that expertise may not develop in a monotonic fashion. Changes in knowledge may not be adequately represented as simple monotonic increases in similarity to some ideal knowledge representation. The results also indicated that the provision of a context when measuring mental model knowledge with different methods may produce conflicting results. Finally, the benefits associated with using performance as the criterion when assessing validity are discussed.