Search processes in answering questions about stories: A possible model

Anderson's (1976) ACT theory was slightly modified and applied to Wh-question answering and to natural prose materials. Assuming Anderson's memory representation of sentences, the hearing or reading of a question activates the memory representations of the content words in the question. Activation fromall the content words spreads in parallel along paths linking those words and the words to the proposition node. Activation from the different content words intersects at a node. After intersection, the labels on the paths leading to the intersection are checked, and if the labels are appropriate, the answer is given. In the case of negative questions, questions whose answer is not told because the answer is not in the story, the situation is more complex. The model generated 10 predictions, for both positive and negative questions, and all predictions were supported. In line with all previous research, negative questions, collapsed over all variables, took longer to respond to than positive questions, also collapsed over all variables. This was interpreted as a more extensive search taking place with negatives.     

Childhood amnesia: on answering questions about very early life events

Twenty five young adults were asked about the events surrounding the birth of a younger sibling which took place when they were under the age of 2 years. Approximately 40% of the participants claimed to have significant memories of the events. The mothers of our participants verified that a majority of their answers were accurate. Comparing the pattern of data with those previously collected (Eacott & Crawley, 1998) suggests that the memories of those who were aged below 2:0 are qualitatively similar to the memories of those who were older at the time of events and dissimilar in type to those who are basing their reports on reconstructions from family knowledge. This finding may be evidence that memories of events that occurred before the age of 2 years are genuine but rare. This conclusion may be useful in assessing theories of childhood amnesia.     

A Matlab function to estimate choice model parameters from paired-comparison data

Tversky (1972) has proposed a family of models for paired-comparison data that generalize the Bradley—Terry—Luce (BTL) model and can, therefore, apply to a diversity of situations in which the BTL model is doomed to fail. In this article, we present a Matlab function that makes it easy to specify any of these general models (EBA, Pretree, or BTL) and to estimate their parameters. The program eliminates the time-consuming task of constructing the likelihood function by hand for every single model. The usage of the program is illustratedby several examples. Features of the algorithm are outlined. The purpose of this article is to facilitate the use of probabilistic choice models in the analysis of data resulting from paired comparisons.     

A Matlab function to estimate choice model parameters from paired-comparison data.

Tversky (1972) has proposed a family of models for paired-comparison data that generalize the Bradley-Terry-Luce (BTL) model and can, therefore, apply to a diversity of situations in which the BTL model is doomed to fail. In this article, we present a Matlab function that makes it easy to specify any of these general models (EBA, Pretree, or BTL) and to estimate their parameters. The program eliminates the time-consuming task of constructing the likelihood function by hand for every single model. The usage of the program is illustrated by several examples. Features of the algorithm are outlined. The purpose of this article is to facilitate the use of probabilistic choice models in the analysis of data resulting from paired comparisons.     

The process of answering direction-giving questions when someone is lost on a university campus: The role of pragmatics

Two experiments investigated how college students answered direction-giving questions when a confederate asked for directions to a destination on a university campus. The experiments applied the QUEST model (Graesser and Franklin, 1990) to direction giving, emphasizing the pragmatic component of the model that focuses on establishing common ground and dealing with the questioner's goals. The two experiments had different articulations of the direction-giving question (i.e. ‘How do you get to destination X?’ versus ‘Where is destination X?’), and a different destination. The answers generated by subjects supported both aspects of the pragmatic component.     

A modified beta binomial model with applications to multiple choice and taste tests

A modified beta binomial model is presented for use in analyzing ramdom guessing multiple choice tests and certain forms of taste tests. Detection probabilities for each item are distributed beta across the population subjects. Properties for the observable distribution of correct responses are derived. Two concepts of true score estimates are presented. One, analogous to Duncan's empirical Bayes posterior mean score, is appropriate for assessing the subject's performance on that particular test. The second is more suitable for predicting outcomes on similar tests.This research was made possible by a grant from the Center for Food Policy Research, Graduate School of Business, Columbia University.     

Neural correlates of feigned memory impairment are distinguishable from answering randomly and answering incorrectly: an fMRI and behavioral study

Previous functional magnetic resonance imaging (fMRI) studies have identified activation in the prefrontal-parietal-sub-cortical circuit during feigned memory impairment when comparing with truthful telling. Here, we used fMRI to determine whether neural activity can differentiate between answering correctly, answering randomly, answering incorrectly, and feigned memory impairment. In this study, 12 healthy subjects underwent block-design fMRI while they performed digit task of forced-choice format under four conditions: answering correctly, answering randomly, answering incorrectly, and simulated feigned memory impairment. There were three main results. First, six areas, including the left prefrontal cortex, the left superior temporal lobe, the right postcentral gyrus, the right superior parietal cortex, the right superior occipital cortex, and the right putamen, were significantly modulated by condition type. Second, for some areas, including the right superior parietal cortex, the right postcentral gyrus, the right superior occipital cortex, and the right putamen, brain activity was significantly greater in feigned memory impairment than answering randomly. Third, for the areas including the left prefrontal cortex and the right putamen, brain activity was significantly greater in feigned memory impairment than answering incorrectly. In contrast, for the left superior temporal lobe, brain activity was significantly greater in answering incorrectly than feigned memory impairment. The results suggest that neural correlates of feigned memory impairment are distinguishable from answering randomly and answering incorrectly in healthy subjects.     

The dynamics of choice among multiple alternatives

We consider neurally based models for decision-making in the presence of noisy incoming data. The two-alternative forced-choice task has been extensively studied, and in that case it is known that mutually inhibited leaky integrators in which leakage and inhibition balance can closely approximate a drift-diffusion process that is the continuum limit of the optimal sequential probability ratio test (SPRT). Here we study the performance of neural integrators in n?2 alternative choice tasks and relate them to a multihypothesis sequential probability ratio test (MSPRT) that is asymptotically optimal in the limit of vanishing error rates. While a simple race model can implement this ‘max-vs-next’ MSPRT, it requires an additional computational layer, while absolute threshold crossing tests do not require such a layer. Race models with absolute thresholds perform relatively poorly, but we show that a balanced leaky accumulator model with an absolute crossing criterion can approximate a ‘max-vs-ave’ test that is intermediate in performance between the absolute and max-vs-next tests. We consider free and fixed time response protocols, and show that the resulting mean reaction times under the former and decision times for fixed accuracy under the latter obey versions of Hick's law in the low error rate range, and we interpret this in terms of information gained. Specifically, we derive relationships of the forms log(n-1), log(n), or log(n+1) depending on error rates, signal-to-noise ratio, and the test itself. We focus on linearized models, but also consider nonlinear effects of neural activities (firing rates) that are bounded below and show how they modify Hick's law.     

Human risky choice under temporal constraints: tests of an energy-budget model

Risk-sensitive foraging models predict that choice between fixed and variable food delays should be influenced by an organism's energy budget. To investigate whether the predictions of these models could be extended to choice in humans, risk sensitivity in 4 adults was investigated under laboratory conditions designed to model positive and negative energy budgets. Subjects chose between fixed and variable trial durations with the same mean value. An energy requirement was modeled by requiring that five trials be completed within a limited time period for points delivered at the end of the period (block of trials) to be exchanged later for money. Manipulating the duration of this time period generated positive and negative earnings budgets (or, alternatively, "time budgets"). Choices were consistent with the predictions of energy-budget models: The fixed-delay option was strongly preferred under positive earnings-budget conditions and the variable-delay option was strongly preferred under negative earnings-budget conditions. Within-block (or trial-by-trial) choices were also frequently consistent with the predictions of a dynamic optimization model, indicating that choice was simultaneously sensitive to the temporal requirements, delays associated with fixed and variable choices on the upcoming trial, cumulative delays within the block of trials, and trial position within a block.     

A model of transitive choice

We report a formal model of transitive inference based on protocols from experiments on squirrel monkeys solving the 5-term series problem (McGonigle & Chalmers, 1977, 1992). These studies generate databases featuring transitive choice, task transfer (where at first a significant decrement is observed, and later substantial improvement without explicit training), and, finally, a Symbolic Distance Effect (SDE) based on decision-time data. Using a rule-based (production) system, we first established rule stacks at the group, then at the individual level, on the basis of triadic transfer performance first recorded in the McGonigle and Chalmers (1977) study. The models for each subject then accommodated data from the more intensive, later study with the same subjects (McGonigle & Chalmers, 1992). We found the initial model capable of dealing with all choice and reaction-time phenomena reported thus far, with only small changes in a rule search procedure. In common with an independent assay by McGonigle and Chalmers (1992), our model-based reassessment of decision times indicates that a major source of reaction time variation is item prominence in the rule stack rather than interitem (ordinal) distance per se.     

A contextual model of concurrent-chains choice

An extension of the generalized matching law incorporating context effects on terminal-link sensitivity is proposed as a quantitative model of behavior under concurrent chains. The contextual choice model makes many of the same qualitative predictions as the delay-reduction hypothesis, and assumes that the crucial contextual variable in concurrent chains is the ratio of average times spent, per reinforcement, in the terminal and initial links; this ratio controls differential effectiveness of terminal-link stimuli as conditioned reinforcers. Ninety-two concurrent-chains data sets from 19 published studies were fitted to the model. Averaged across all studies, the model accounted for 90% of the variance in pigeons' relative initial-link responding. The model therefore demonstrates that a matching law analysis of concurrent chains—the assumption that relative initial-link responding equals relative terminal-link value—remains quantitatively viable. Because the model reduces to the generalized matching law when terminal-link duration is zero, it provides a quantitative integration of concurrent schedules and concurrent chains.