Bootstrapping and knowledge of reliability

This is a critical discussion of a paper on the problem of bootstrapping by Jose Zalabardo.

Bootstrapping our way to samesaying

This paper articulates two constraints on an acceptable account of meaning: (i) accessibility: sameness of meaning affords an immediate appearance of de jure co-reference, (ii) flexibility: sameness of meaning tolerates open-ended variation in speakers?? substantive understanding of the reference. Traditional accounts of meaning have trouble simultaneously satisfying both constraints. I suggest that relationally individuated meanings provide a promising way of avoiding this tension. On relational accounts, we bootstrap our way to de jure co-reference: the subjective appearance of de jure co-reference helps make it the case that two token representations really do co-refer.     

Bootstrapping in miniature language acquisition

Given the difficulties in learning meanings of words by observing the referent, it has been suggested that children use the syntactic context of the word to predict part of its meaning, a hypothesis known as syntactic bootstrapping. Semantic bootstrapping is the opposite theory that the knowledge of semantics helps in acquiring syntax. While there is evidence that children can apply their knowledge of correlations between syntax and semantics to perform bootstrapping, it is not clear how they come to know about these correlations in the first place. Here, a connectionist network is presented that learns to comprehend a miniature language by associating sentences with the corresponding scenes. In doing so, it learns the syntactic/semantic correlations and exhibits bootstrapping behavior. It is argued that such specialized phenomena can emerge when general mechanisms are applied to a specific task, and it is not always necessary to endow the learner with pre-existing specialized mechanisms.     

The Normative Requirement of Means-End Rationality and Modest Bootstrapping

“Myth theorists” have recently called the normative requirement of means-end rationality into question. I show that we can accept certain lessons from the Myth Theorists and also salvage our intuition that there is a normative requirement of means-end rationality. I argue that any appeal to a requirement to make our attitudes coherent as such is superfluous and unnecessary in order to vindicate the requirement of means-end rationality and also avoid the problematic conclusion that persons ought to take the means to whatever ends they happen to intend.     

Bootstrapping cognition from behavior-a computerized thought experiment

We show that simple perceptual competences can emerge from an internal simulation of action effects and are thus grounded in behavior. A simulated agent learns to distinguish between dead ends and corridors without the necessity to represent these concepts in the sensory domain. Initially, the agent is only endowed with a simple value system and the means to extract low-level features from an image. In the interaction with the environment, it acquires a visuo-tactile forward model that allows the agent to predict how the visual input is changing under its movements, and whether movements will lead to a collision. From short-term predictions based on the forward model, the agent learns an inverse model. The inverse model in turn produces suggestions about which actions should be simulated in long-term predictions, and long-term predictions eventually give rise to the perceptual ability.     

Bootstrapping the mind: analogical processes and symbol systems

Human cognition is striking in its brilliance and its adaptability. How do we get that way? How do we move from the nearly helpless state of infants to the cognitive proficiency that characterizes adults? In this paper I argue, first, that analogical ability is the key factor in our prodigious capacity, and, second, that possession of a symbol system is crucial to the full expression of analogical ability.     

A Bootstrapping Model of Frequency and Context Effects in Word Learning

Prior research has shown that people can learn many nouns (i.e., word–object mappings) from a short series of ambiguous situations containing multiple words and objects. For successful cross‐situational learning, people must approximately track which words and referents co‐occur most frequently. This study investigates the effects of allowing some word‐referent pairs to appear more frequently than others, as is true in real‐world learning environments. Surprisingly, high‐frequency pairs are not always learned better, but can also boost learning of other pairs. Using a recent associative model (Kachergis, Yu, & Shiffrin, 2012), we explain how mixing pairs of different frequencies can bootstrap late learning of the low‐frequency pairs based on early learning of higher frequency pairs. We also manipulate contextual diversity, the number of pairs a given pair appears with across training, since it is naturalistically confounded with frequency. The associative model has competing familiarity and uncertainty biases, and their interaction is able to capture the individual and combined effects of frequency and contextual diversity on human learning. Two other recent word‐learning models do not account for the behavioral findings.     

Parts and differences

Part/whole is said in many ways: the leg is part of the table, the subset is part of the set, rectangularity is part of squareness, and so on. Do the various flavors of part/whole have anything in common? They may be partial orders, but so are lots of non-mereological relations. I propose an “upward difference transmission” principle: x is part of y if and only if x cannot change in specified respects while y stays the same in those respects.     

Using the Bollen-Stine Bootstrapping Method for Evaluating Approximate Fit Indices

Accepting that a model will not exactly fit any empirical data, global approximate fit indices quantify the degree of misfit. Recent research (Chen, Curran, Bollen, Kirby, & Paxton, 2008) has shown that using fixed conventional cut-points for approximate fit indices can lead to decision errors. Instead of using fixed cut points for evaluating approximate fit indices, this study focuses on the meaning of approximate fit and introduces a new method to evaluate approximate fit indices. Millsap (2012) introduced a simulation-based method to evaluate approximate fit indices. A limitation of Millsap's (2012) work was that a rather strong assumption of multivariate normality was implied in generating simulation data. In this study, the Bollen-Stine bootstrapping procedure (Bollen & Stine, 1993) is proposed to supplement the former study. When data are nonnormal, the conclusions derived from Millsap's (2012) simulation method and the Bollen-Stine method can differ. Examples are given to illustrate the use of the Bollen-Stine bootstrapping procedure for evaluating the Root Mean Squared Error of Approximation (RMSEA). Comparisons are made with the simulation method. The results are discussed, and suggestions are given for the use of proposed method.     

Bootstrapping conceptual deduction using physical connection: rethinking frontal cortex

The age at which infants can demonstrate the ability to deduce abstract rules can be reduced by more than half, from 21 months to 9 months. The key is to introduce a physical connection between the items to be conceptually related. I argue here that making the same change in how items are presented might also help some preschoolers with learning delays, especially some children with autism. I also suggest that the roles of premotor and ventrolateral prefrontal cortices in deducing abstract rules might have been misinterpreted behaviorally and anatomically. The crucial brain region may be the periarcuate, which partially overlaps both premotor and lateral prefrontal cortex. The cognitive ability made possible by this region might be something far more elementary than previously considered: the ability to perceive conceptual connections in the absence of physical connection.     

Bootstrapping the lexicon: a computational model of infant speech segmentation

Prelinguistic infants must find a way to isolate meaningful chunks from the continuous streams of speech that they hear. BootLex, a new model which uses distributional cues to build a lexicon, demonstrates how much can be accomplished using this single source of information. This conceptually simple probabilistic algorithm achieves significant segmentation results on various kinds of language corpora - English, Japanese, and Spanish; child- and adult-directed speech, and written texts; and several variations in coding structure - and reveals which statistical characteristics of the input have an influence on segmentation performance. BootLex is then compared, quantitatively and qualitatively, with three other groups of computational models of the same infant segmentation process, paying particular attention to functional characteristics of the models and their similarity to human cognition. Commonalities and contrasts among the models are discussed, as well as their implications both for theories of the cognitive problem of segmentation itself, and for the general enterprise of computational cognitive modeling.     

Parts and boundaries.

Within the framework of Conceptual Semantics, a family of conceptual features and functions is developed that accounts for phenomena in the semantics of noun phrases such as the mass-count distinction, plurality, the partitive construction (a leg of the table), the constitutive construction (a house of wood), the "Universal Packager" (three coffees), and boundary words such as end, edge, and crust. Using the strong formal parallelism between noun phrase semantics and event structure that is a hallmark of the Conceptual Semantics approach, the features and functions of the NP system are applied to a wide range of problems in event structure, for example the analysis of the Vendler classes, the meaning of the progressive, the "imperfective paradox", and "aktionsarten" such as the syntactically unexpressed sense of repetition in The light flashed until dawn. Crucial to the analysis is that these features and functions can be expressed in syntactic structure either by being part of lexical conceptual structure, or by use of a morphological affix, or by being associated with the meaning of a construction such as N of NP or nominal compounding. Alternatively, they may remain unexpressed altogether, being introduced into the conceptual structure of a phrase by "rules of construal". This shows that lexical semantics and phrasal semantics interpenetrate deeply, and that there is no strict one-to-one correspondence between syntactic and semantic structures. In addition, the analysis provides further evidence that natural language semantics must be based on a psychological view of meaning--it must be concerned with how language users are constructed to understand and schematize the world.     

Bootstrapping word order in prelexical infants: a Japanese-Italian cross-linguistic study

Learning word order is one of the earliest feats infants accomplish during language acquisition [Brown, R. (1973). A first language: The early stages, Cambridge, MA: Harvard University Press.]. Two theories have been proposed to account for this fact. Constructivist/lexicalist theories [Tomasello, M. (2000). Do young children have adult syntactic competence? Cognition, 74(3), 209-253.] argue that word order is learned separately for each lexical item or construction. Generativist theories [Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press.], on the other hand, claim that word order is an abstract and general property, determined from the input independently of individual words. Here, we show that eight-month-old Japanese and Italian infants have opposite order preferences in an artificial grammar experiment, mirroring the opposite word orders of their respective native languages. This suggests that infants possess some representation of word order prelexically, arguing for the generativist view. We propose a frequency-based bootstrapping mechanism to account for our results, arguing that infants might build this representation by tracking the order of functors and content words, identified through their different frequency distributions. We investigate frequency and word order patterns in infant-directed Japanese and Italian corpora to support this claim.     

Resampling approach to statistical inference: Bootstrapping from event-related potentials data

We propose the use of the bootstrap resampling technique as a tool to assess the within-subject reliability of experimental modulation effects on event-related potentials (ERPs). The assessment of the within-subject reliability is relevant in all those cases when the subject score is obtained by some estimation procedure, such as averaging. In these cases, possible deviations from the assumptions on which the estimation procedure relies may lead to severely biased results and, consequently, to incorrect functional inferences. In this study, we applied bootstrap analysis to data from an experiment aimed at investigating the relationship between ERPs and memory processes. ERPs were recorded from two groups of subjects engaged in a recognition memory task. During the study phase, subjects in Group A were required to make an orthographic judgment on 160 visually presented words, whereas subjects in Group B were only required to pay attention to the words. During the test phase all subjects were presented with the 160 previously studied words along with 160 new words and were required to decide whether the current word was “old” or “new.” To assess the effect of word imagery value, half of the words had a high imagery value and half a low imagery value. Analyses of variance performed on ERPs showed that an imagery-induced modulation of the old/new effect was evident only for subjects who were not engaged in the orthographic task during the study phase. This result supports the hypothesis that this modulation is due to some aspect of the recognition memory process and not to the stimulus encoding operations that occur during the recognition memory task. However, bootstrap analysis on the same data showed that the old/new effect on ERPs was not reliable for all the subjects. This result suggests that only a cautious inference can be made from these data.     

Bootstrapping to test for nonzero population correlation coefficients using univariate sampling

This article proposes 2 new approaches to test a nonzero population correlation (rho): the hypothesis-imposed univariate sampling bootstrap (HI) and the observed-imposed univariate sampling bootstrap (OI). The authors simulated correlated populations with various combinations of normal and skewed variates. With alpha set=.05, N> or =10, and rho< or =0.4, empirical Type I error rates of the parametric r and the conventional bivariate sampling bootstrap reached .168 and .081, respectively, whereas the largest error rates of the HI and the OI were .079 and .062. On the basis of these results, the authors suggest that the OI is preferable in alpha control to parametric approaches if the researcher believes the population is nonnormal and wishes to test for nonzero rhos of moderate size.