Experiments and models in sequential integration of information

Using the example of bimodal temporal order judgments, we investigate and explain the transition from static designs mapping single stimulus presentations to single responses to designs which require some kind of integration of the processed information. First, we estimate the parameters of several well-known channel-oriented models (Baron, 1969; Gibbon & Rutschmann, 1969) for temporal order judgments. Subsequently, we generalize these models to explain the results of two more complex experimental designs and derive and test their predictions. Special concern is given to the question whether or not the models parameters remain invariant across the experiments. The basic assumptions of the channel-oriented models are confirmed by our data, they represent a well-suited platform to explain more complex integrative processes. Within a sequential decision design, where subjects are allowed to observe a stimulus arbitrarily often before their terminal decision, the individual representations of the stimuli are additively integrated. Hence, even information, which did not yet lead to a terminal decision, will generally influence the later outcome. This representation of the stimuli is accompanied by a dynamic decision rule, which is clearly time-dependent. Several too simplizistic alternative models can be rejected if (and only if) one derives their predictions in some detail and tests them with sufficient statistical power.     

The Morton-Massaro law of information integration: implications for models of perception

Information integration may be studied by analyzing the effect of 2 or more sources (e.g., auditory and visual) on participants' responses. Experiments show that ratios of response probabilities often factorize into components selectively influenced by only 1 source (e.g., 1 component affected by the acoustic source and another 1 affected by the visual source). This is called the Morton-Massaro law (MML). This article identifies conditions in which the law is optimal and notes that it reflects an implicit assumption about the statistics of the environment. Adherence to the MML can be used to assess whether the assumption is being made, and analyses of natural stimuli can be used to determine whether the assumption is reasonable. Feed-forward and interactive models subject to a channel separability constraint are consistent with the law.     

The integration of spatial information across different viewpoints

The integration of spatial information perceived from different viewpoints is a frequent, yet largely unexplored, cognitive ability. In two experiments, participants saw two presentations, each consisting of three targets—that is, illuminated tiles on the floor—before walking the shortest possible path across all targets. In Experiment 1, participants viewed the targets either from the same viewpoint or from different viewpoints. Errors in recalling targets increased if participants changed their viewpoints between presentations, suggesting that memory acquired from different viewpoints had to be aligned for integration. Furthermore, the error pattern indicates that memory for the first presentation was transformed into the reference frame of the second presentation. In Experiment 2, we examined whether this transformation occurred because new information was integrated already during encoding or because memorized information was integrated when required. Results suggest that the latter is the case. This might serve as a strategy for avoiding additional alignments.     

Misleading postevent information: Testing parameterized models of integration in memory

A classic series of experiments by Loftus, Miller and Burns (1978) showed that a person's recollection of an event can be changed by misleading postevent information. Several hypotheses accounting for this effect have been proposed. Loftus' hypothesis of destructive updating claims that the original memory is destroyed by the postevent information. The coexistence hypothesis asserts that the older memory survives but is rendered inaccessible through a mechanism of inhibition or suppression. The non-conflict hypothesis simply accounts for the effect by claiming that subjects can only be misled if they did not encode or if they forgot the original event. These three hypotheses were modelled with the help of all-or-none probabilistic event trees. An experiment was conducted in order to test the three models and to assess parameter values. The experiment followed the classic Loftus paradigm. We suggested to some subjects that they had seen a stopsign, whereas in fact they had seen a traffic light. The misleading postevent information resulted in poorer reproduction of traffic light. Later, all subjects were asked whether they could remember the color of the traffic light, even if they believed they had seen a stopsign. The results showed that subjects who received the misleading post-event information were at least as good at recalling the color of the traffic light as subjects who did not receive misleading information. The no-conflict model accounts well for the obtained results, although the two other, less parsimonious, models cannot be entirely rejected.     

A new interpretation of stochastic test models

A new look at latent trait models is proposed. The event of an item being solved by a person is related to the event that the momentary value of a person-specific random component is at least as large as the corresponding value of an item-specific random component. The Birnbaum logistic test model is shown to be generated by a bivariate extreme value distribution for the components. Some consequences of this interpretation are outlined.I am indebted to David Strauss for calling the extreme value distribution (1) to my attention in a random utility context. The paper also benefited from discussions with H. C. Micko, H. H. Schulze and K. F. Wender.     

A unifying model for the structure of intellectual abilities

Models of the structure of cognitive abilities suggested by Spearman, Thurstone, Guilford, Vernon and Cattell-Horn are reviewed. It is noted that some of the models include a general intellectual factor (g) while others do not. It is also noted that some models are nonhierarchical, while in others more narrow abilities are subsumed under broader abilities in a hierarchical pattern. An empirical study in which a test battery of 16 tests was administered to some 1000 subjects in the 6th grade is reported. Using the LISREL technique to test different models, good support is obtained for oblique primary factors in the Thurstone tradition as well as for the second-order factors fluid intelligence, crystallized intelligence, and general visualization hypothesized by Cattell and Horn. It is also found, however, that the second-order factor of fluid intelligence i is identical with a third-order g-factor. On the basis of these results a three-level model (the HILI-model) is suggested, with the g-factor at the top, two broad factors reflecting the ability to deal with verbal and figural information, respectively, at the second-order level, and the primary factors in the Thurstone and Guilford tradition at the lowest level. It is argued that most previously suggested models are special cases of the HILI-model.     

Invariance of inference time when information was presented in different linguistic formats

Ss read paragraphs varying in syntactic complexity, contiguity of information that was conceptually related, and amount of information relating to quantitative variables. Paragraphs were presented too briefly to allow full reading in many cases. In questions requiring use of the quantitative information, proportion of correct responses was influenced by syntax, contiguity, and number of variables from the paragraph. However, neither syntax nor contiguity influenced latency of correct answers, while number of variables, a factor affecting complexity of the process of answering, did influence correct response latency. The findings confirm and extend earlier results by Kintsch and Monk (1972) and give further support to the proposition that assimilation of information in linguistic input involves storing the semantic content of the message.     

Alexithymia and the interpretation of emotion-relevant information

Abstract

This study examined the relationship between alexithymia and (a) hemis-patial bias; (b) the accuracy of interpreting emotion-relevant information; and (c) the style of interpreting emotion-relevant information. Subjects were 137 college students. Alexithymia was measured using the Toronto Alexithymia Scale. Hemispatial bias was measured using a free-vision chimeric face task. Interpretations of emotion-relevant information were assessed using the Profile of Nonverbal Sensitivity (PONS) and a second instrument designed for this study called the Emotional Story Test (EST). Although high levels of alexithymia were associated with hemispatial bias, alexithymia was not associated with the accuracy of subjects' interpretations. In contrast, alexithymia was associated with low ratings of emotional intensity and with subjects' inclinations to choose angry and dominant interpretations.     

Expectation theories of decision making for duplex gambles

Complete tests of subjectively expected utility (SEU), subjectively expected value (SEV), expected utility (EU) and expected value (EV) theories were made for duplex gambles without measuring subjective probability or subjective utility. All gambles were hypothetical and offered on booklets. The duplex gambles consisted of winning gambles, which offered a chance to win a certain amount of money or to break even; and losing gambles, which offered a chance to lose a certain amount of money or break even.The results indicated that SEU, SEV and EU theories could not account for the strategies of 33%, 53% and 86% of the Ss respectively in the losing form of gambles, while EV theory accounted for 78% of the behavior of Ss.In the winning form of gambles, SEU, SEV and EU theory held for 77%, 65%, and 54% of the Ss respectively, while EV theory held for only 40% of the Ss. Suggestions for further research were made.     

Bayesian integration of spatial information

Spatial judgments and actions are often based on multiple cues. The authors review a multitude of phenomena on the integration of spatial cues in diverse species to consider how nearly optimally animals combine the cues. Under the banner of Bayesian perception, cues are sometimes combined and weighted in a near optimal fashion. In other instances when cues are combined, how optimal the integration is might be unclear. Only 1 cue may be relied on, or cues may seem to compete with one another. The authors attempt to bring some order to the diversity by taking into account the subjective discrepancy in the dictates of multiple cues. When cues are too discrepant, it may be best to rely on 1 cue source. When cues are not too discrepant, it may be advantageous to combine cues. Such a dual principle provides an extended Bayesian framework for understanding the functional reasons for the integration of spatial cues.     

A unifying expression for the maximal reliability of a linear composite

A formally simple expression for the maximal reliability of a linear composite is provided, its theoretical implications and its relation to existing results for reliability are discussed.     

The integration of motivational information: A conjoint measurement analysis

Within Anderson's (1974a, b, c, 1978b) information integration theory, the integration of motivational information was investigated by means of conjoint measurement techniques. Eighteen university students were asked to judge hypothetical co-students characterized by three features (intelligence, motivation and the extent to which they study) according to their chances to pass. Both rank order data and ratings were obtained. The orderings of most subjects could be represented very well by an additive model. A polynomial regression procedure was applied to determine the shape of the response function for the ratings. As this function was quite linear for all subjects, the ratings could be said to form an interval scale.     

MMPI-Rorschach interrelationships: proposed criteria for evaluating explanatory models

An extensive empirical literature, spanning 50 years and 45 published investigations, leads to the conclusion that the Rorschach and the Minnesota Multiphasic Personality Inventory (MMPI) bear little or no meaningful relationship to each other. An inherent tension exists between these empirical findings and the widespread clinical practice of integrating MMPI and Rorschach results. At least three perspectives have been advanced concerning the relationship between these two instruments. One view holds that variables from the two measures will demonstrate significant patterns of convergence only in carefully designed research studies that involve specific, well-constructed and theoretically derived predictions. A second view postulates that although variables from the Rorschach and MMPI do not bear a high intercorrelation with each other, these variables may be combined to contribute significantly to the prediction of outcome variance. In this view, the combined use of the two instruments yields clinically useful increases in incremental validity not achievable by the use of either instrument in isolation. A third approach is reflected in the proposition that although the MMPI and Rorschach do not produce significant interrelationships under general conditions, meaningful relationships may be yielded under specific psychometric conditions, for example, under conditions in which the response styles are consistently displayed across those two instruments. This article will review research evidence related to each of these perspectives, with particular emphasis on the importance of clearly specifying the criteria to be applied in evaluating these models. The concepts of parsimony and heuristic value are proposed for these evaluation purposes.     

Breaking the rules in perceptual information integration

We develop a broad theoretical framework for modelling difficult perceptual information integration tasks under different decision rules. The framework allows us to compare coactive architectures, which combine information before it enters the decision process, with parallel architectures, where logical rules combine independent decisions made about each perceptual source. For both architectures we test the novel hypothesis that participants break the decision rules on some trials, making a response based on only one stimulus even though task instructions require them to consider both. Our models take account of not only the decisions made but also the distribution of the time that it takes to make them, providing an account of speed-accuracy tradeoffs and response biases occurring when one response is required more often than another. We also test a second novel hypothesis, that the nature of the decision rule changes the evidence on which choices are based. We apply the models to data from a perceptual integration task with near threshold stimuli under two different decision rules. The coactive architecture was clearly rejected in favor of logical-rules. The logical-rule models were shown to provide an accurate account of all aspects of the data, but only when they allow for response bias and the possibility for subjects to break those rules. We discuss how our framework can be applied more broadly, and its relationship to Townsend and Nozawa’s (1995) Systems-Factorial Technology.     

Invariance of the weight parameter in information integration

The relative effect of each informational dimension in an information integration task is a joint function of its weight and the range of values over which it is varied. A method is developed for separating these two factors. Weights obtained by this method were compared across variations of stimulus range. Subjects rated the performance of students on the basis of midterm exam scores and final exam scores. For some subjects, the range of scores on the final exam was twice that on the midterm and the reverse was true for other subjects. An averaging model was shown to describe the results, and weights did not differ for different stimulus ranges. This was true for each of two instructional conditions: one in which a particular weighting strategy was prescribed and one in which there was no prescribed weighting. Students who were instructed to use a 2∶1 weighting were found to assign more than twice as much weight to the final as to the midterm.