No description available.

No description available.

The experiments reported in this thesis attempted to directly study the process of look-ahead during problem solving. Recent work has suggested that interface manipulations that increase look-ahead during problem solving lead to improvements in performance. However, evidence has been indirect, and there have been few attempts specifically made to quantify look-ahead span, changes that may occur over time and possible interactions with the task environment. An initial experiment required users to specify 3 moves in advance while solving the 8-puzzle. The strict enforcing of look-ahead by even a small number of moves was unsuccessful in terms of improving problem performance. In fact, results indicated that such move enforcement may negatively affect performance. Subsequent experiments, using both the 8-puzzle and Water Jars problems, provided participants with a motivation to plan using a Scoreboard system that rewarded greater planning and look-ahead. Results found this approach to be more viable, as the interface appeared to support the opportunistic planning behaviour frequently undertaken by participants. Across a series of experiments, increased look-ahead led to more efficient problem solving performance compared to controls, while leaving total time to solution unaffected. Look-ahead span increased to approximately 11 steps when transforming the same start-state to a goal-state over trials on the 8-puzzle. When a new solution path had to be generated for each new problem start-state, look-ahead still increased over trials, but only to a span of approximately 4 steps. This look-ahead span was also observed during Water Jars performance when the Scoreboard manipulation was present. A manipulation of 'system response time' (SRT) on Water Jars problems also led to improved performance but indicated an adaptation to the manipulation, leading to a lesser impact of SRT than previous manipulations. The results are discussed in relation to existing studies of planning, performance and the role that look-ahead may have in future studies of problem solving.

The present thesis examined the processes responsible for strategy selection in problem-solving tasks. Despite the salience of this mechanism there has been a dearth in empirical research in the paradigm. Existing accounts, primarily modelled upon simulations of data sets, propose that strategies are not selected per se, but problems are solved by an automatic attempt to retrieve a solution (the Automaticity account Logan, 1988 2002). Contrary to this account four studies presented in the first empirical series demonstrated that predicted retrieve/calculate selections could be made rapidly (within 850 ms) and accurately. This indicated that problem-solving comprises two dissociable phases, selection then solution. Selection was found to be sensitive to the familiarity of a problem and also specific problem features supporting an account in which selection may be determined by the type of problem and the context in which the problem is solved (the Adaptive account Reder & Ritter, 1992 Siegler & Araya, 2005). Elucidating the mechanisms responsible for these effects, in the second empirical series, three issues representative of real-world problem-solving episodes were examined. When multiple cues to selection were available, the interplay between the cues either served to inhibit the effects of both cues, or facilitated the effect of one cue at the expense of the other. Problem familiarity effects were attributed to implicit procedures as these effects were apparently un-reliant upon conscious processes (Reder & Ritter, 1992 Schunn et al, 1997). However, the feature identification process, rather than the selection mechanism itself, was found to be reliant upon consciously directed processes (Siegler & Araya, 2005). The findings from these studies were used to evaluate existing accounts of strategy selection, and reflecting limitations in these models, candidate mechanisms are proposed to account for the key effects revealed in this thesis.

The main focus of this thesis was to design training to mitigate the effects of constraints underlying verbal insight problem solving. Concurrent verbal protocols were collected. Experiments 1 and 2 tested training that was specific to solving problems that were exemplars of trained categories. In Experiment 1, heuristic training was provided for two categories of constraint: those with ambiguous words and those with human names that should be associated with animals. Transfer was positive for novel problems within the two trained categories but not for problems outside. Experiment 2 improved performance on problems with ambiguous words once shortcomings in Experiment 1 were addressed. Experiment 3 tested training that was specific to solving functional fixedness verbal insight problems. An iterative process of considering many functions of individual items was successful in facilitating performance but not for problems outside. Experiments 4 and 5 investigated the effectiveness of different types of generic training in facilitating solution of novel verbal insight problems. In Experiment 4, participants were trained to identify assumptions, which they made during problem solving and were inconsistent with the problem statement. In Experiment 5, participants were trained to iteratively consider different parts of the problem specification and to identify inconsistencies as in Experiment 4. Solution rate was improved in both experiments although instruction to explain and justify oneself during problem solving was also sufficient in facilitating performance in Experiment 5. Finally, Experiment 6 tested a novel method for identifying when a participant was constrained by an incorrect representation during verbal insight problem solving. The results supported that there is variability in the nature of stereotypical constraints involved and demonstrated how training can be designed to induce restructuring or a shift in representation.

No description available.

No description available.

The discipline of Critical Thinking is a methodology for the analysis and appraisal of natural language argument, commonly referred to as ‘real’ argument, a term intended to signify a contrast between the reasoning found in ordinary written or spoken texts, and the specialised constructions tailored by logicians to exemplify validity. There are, accordingly, two perspectives from which to view arguments: the arguer’s and the critic’s. From the critic’s perspective an argument is simply an object of appraisal. From the arguer’s perspective, however, an argument is more intimately associated with the act or practice of arguing or, as many commentators prefer to say, an argument is a product of the practice (of argumentation). There is hence an ambiguity between the act and object senses of ‘argument’, as well as a difference between object-senses themselves, depending on the perspective taken. For the traditional logician, an argument is a set of sentences or propositions, one of which is claimed, supposed, or intended, not necessarily in any active sense, to follow from the other(s). Such a conception facilitates assessment on the criterion of validity. Indeed, an argument can be defined as an object that is either valid or invalid. For the critic of ‘real’ argument, the object is what is propounded by a speaker or author, typically for persuasive purposes. Arguments from this perspective are not mere sets of arbitrarily designated propositions, but claims and inferences identified by the critic based on the interpretation and classification of actual texts. The aim of the thesis is to integrate these two conceptions of argument by identifying a point of intersection between the two perspectives: that is, to explain how the object of acts of argument can be seen to coincide with the objects of logical appraisal. I argue that the act of propounding an argument is essentially an assertive act, its object a complex proposition. It is more precise, however, to see propounding an argument as two mutually complementary acts, also assertives: 1) premising (or reason-giving) and 2) concluding (inferring). Premising is directly assertive, its objects the premises themselves. Concluding, however, is more than just the assertion of a conclusion. If it were not, the argument would merely be a conjunction. What is asserted, in an act of concluding, is equivalent to a conditional formed from the conjunction of premises, (A), as antecedent, and the conclusion, (C), as consequent. What is asserted, then, in the whole act of propounding an argument, is the conjunctive proposition: (1) A and (if A then C) An assertive utterance of (1) commits the speaker to C by modus ponens. Hence, I propose, the act of argument (i.e., the propounding of an argument) is naturally deductive, and its object a deductive argument accordingly. By the same token its object (i.e., that which is propounded) corresponds with the ordered and indexed set designated by the logician as an object of appraisal. This account, I conclude, has important implications for critical thinking. It provides a theoretical basis, a groundwork, on which to develop a deductivist methodology for appraising natural-language arguments. Whilst I do not go into detail on the practical application of natural-language deductivism (NLD), I take the thesis to be an endorsement, and justification, for its implementation, and for a greater role within critical thinking for logic.

No description available.

This thesis explores conflict in reasoning using a mouse-tracking paradigm that measures participants' instantaneous attraction towards competing response options. It focuses on two kinds of conflict: between competing sources of information in inductive reasoning, and between "Type 1" and "Type 2" processes in Dual Process accounts of reasoning. The mouse-tracking data reveal under what circumstances conflict occurs, at what points in time participants are influenced by different factors, and something of the qualitative nature of this conflict.