Access to grammatical gender in Hebrew was examined using gender decisions, lexical decision, two-word lexical decision, and grammaticality judgments (with noun-adjective pairs, noun-verb phrases, and plural noun-plural adjective pairs). In the gender decision task, the role of word-form was dramatic. Nouns with an explicit gender marker (regular feminine) were classified most easily, and next were nouns in the default gender (unmarked masculine). In contrast, exception nouns (unmarked feminine nouns) produced extremely slow RTs and unusually high error rates. These same exception nouns, however, produced radically different results when syntactic context was provided; in the detection of gender agreement (i.e., "yes" decisions in grammaticality judgments) exception nouns did not produce longer RTs, and also error rates were not higher than those for regular nouns. In contrast, there was a strong effect of regularity in the detection of gender agreement violation (i.e., "no" decisions in grammaticality judgments). This same pattern replicated even when all elements of the phrase were explicitly marked (i.e., plurals grammaticality judgments), and also in the context of a task that placed both grammatical and ungrammatical pairs in the context of the same decision type. Because comprehension tasks necessarily begin with the analysis of word-forms where gender-marking regularity is most conspicuous, it is suggested that the results provide very strong evidence in favor of modular access mechanisms to lemma and lexeme representations. It is further suggested that access to the mental lexicon does not provide conscious access to grammatical gender. Efficient access is available only when the syntactic processor is invoked, and this processor does not rely on regularity at the word-form level. Without a syntactic context grammatical gender is retrieved through a variety of strategies and heuristics. Finally, it appears that gender-marking regularity does play a role in post-lexical processing mechanisms that are designed to detect grammaticality violations, and the latter mechanisms, as well as syntactic processing itself, cannot be suppressed.

The work presented here investigated how word forms are stored and accessed for language production. While the study of single word reading has made significant use of the concept of lexical neighborhoods (the number of similar words there are in the language), the study of word production has not. Data from natural and experimental investigations of both tip-of-the-tongue (TOT) states and word substitution errors were used to evaluate the organizational system for word forms. This type of detailed investigation should aid in discovering the relevant form parameters for determining similarity for production. It was shown that although the parameters of form that are relevant for TOT states and word substitution errors are similar, substitution errors provide a better source of data regarding lexical retrieval. This is because TOT data reflect a variety of problem solving mechanisms which likely do not play a role in lexical retrieval under normal circumstances. The evaluation of specific form parameters overlapping between substitution errors and targets suggested significantly greater degree of similarity than had been reported previously. Furthermore, experimental investigation of word form overlap indicated that the method employed here and elsewhere does not tap the same processes as the natural data. Experimental investigations of TOT states for both real and novel targets suggested that new techniques employed in this dissertation may be useful in evaluating lexical parameters involved in TOT states and in the process of lexicalization in adults.

By examining psittacine speech (primarily from a Grey parrot, Psittacus erithacus, named Alex) in a series of two acoustic studies, two articulatory studies and one perceptual study, this dissertation demonstrates that some aspects of human language are not unique to our species. The first two studies identify frequency, intensity and durational aspects of, respectively, Alex's vowels (/i,I,e,ε,æ,ɒ,ə,o,U,u/) and stop consonants (/p,b,t,d,k,g/) that either differ from or resemble their human counterparts (primarily I. M. Pepperberg). Our results indicate that Alex produces acoustically distinct phonemes using more high frequency information than do humans. In both studies, we use acoustic data to make predictions about articulation. We also compare Alex's speech to that of a mynah, and conclude that these species use different mechanisms to produce speech. The third and fourth study examine vowel articulation: The third study, an X-ray videotape analysis, establishes that Alex configures his vocal tract in distinct ways for /i/ and /ɒ/. The fourth study models more than 2800 Grey parrot vocal tract shapes as conjoined tubes with known area functions and calculates associated vowel formants. The relationship between these mathematical models and formant values for /i,I,e,ε,æ,ɒ,ə,o,U,u/ is consistent with findings from the vowel study, X-ray study, Grese's unpubl. data and personal observations of relationships between Alex's vocal tract configurations and vowel production. The third and fourth studies show that Alex's "phonemes" are articulatorily distinct. The fifth study is perceptual and examines the relationship between acoustic characteristics of psittacine vowels and the accuracy with which listeners perceive them. We find evidence that Alex's acoustically least "human" vowel, /i/, is very difficult for listeners to perceive unless they have substantial exposure to Alex (on the order of several months). More acoustically prototypical vowels, like /ɒ/, are accurately identified even by less experienced listeners. We thus find that, at least for experienced listeners, Alex's "phonemes" are perceptually distinct. In sum, the dissertation provides acoustic, articulatory and perceptual evidence of phonemes and other phonetic structure in the speech of a psittacid.

Underawareness of memory impairment in Alzheimer's disease (AD) was examined using three experimental methodologies: the feeling-of-knowing (FOK) paradigm, the performance prediction/postdiction paradigm (PPP), and patient-caregiver discrepancy on questionnaire (QD). For the FOK paradigm, thirty-two AD patients and their spouses were given an episodic (sentence) memory task, during which they were asked to recall the sentences' final words and to make retrospective confidence judgements about their recall attempts. For failed items, participants rated their future likelihood of correctly recognizing each ending (FOK). Participants' ratings were compared to actual recall or recognition scores. Results revealed that AD patients were less accurate than their non-demented spouses in making retrospective recall and prospective FOK ratings. Similarly, results from the PPP showed that AD patients overestimated their own performance relative to their caregivers' performance on a list learning and memory task, more than caregivers did. Contrary to prior research, participants' responses on questionnaires concerning their own and their spouses' memory change showed that AD patients reported less memory change for both themselves and their caregivers. The paradigms were hypothesized to differ in the extent to which they require different aspects of metamemory ability. Because both require on-line memory monitoring ability, it was hypothesized that the FOK and PPP would be most related, while the QD was predicted to be relatively independent, relying more on generalized self-memory beliefs. A principal components analysis confirmed that the questionnaire data was relatively independent from the other two methodologies. However, the prospective FOK ratings loaded on a different component than the retrospective ratings and the PPP variables. Memory monitoring may be dissociable by task, with prospective ratings relying more on inferential processes, thought to depend on frontal lobe functioning. In order to examine the relationship between underawareness and risky behavior, caregivers were administered a novel interview during which they rated the patients' propensity to attempt various activities, as well as the riskiness of each behavior. Contrary to expectation, riskiness was not significantly correlated with underawareness, but was associated with greater impairments of activities of daily living.

Children can learn aspects of the meaning of a new word on the basis of only a few incidental exposures and can retain this knowledge for a long time. The process of rapidly learning and remembering new words has come to be known as fast mapping. It is often maintained that fast mapping is the result of a dedicated language mechanism, but it is possible that this same capacity might apply in domains other than language learning. The present studies explore the nature of fast mapping, with the goal of revealing more about the mechanism underlying word learning in children. One possibility is that the capacity for word learning is mediated by a specialized language mechanism. A second view posits that all of language acquisition depends on more general cognitive processes. Alternatively, the acquisition of words and grammar may involve different mechanisms. To test these alternative proposals, children and adults were taught a new object name and an arbitrary fact about a novel object, and were tested for their retention immediately or after a delay. The findings revealed that fast mapping is not limited to word learning, suggesting that the capacity to learn and retain new words is the result of learning and memory abilities that are not specific to language. Three further studies then explored the specificity and development of the capacity for fast mapping. The implications of these findings for theories of word learning are discussed.

Almost nothing is known regarding infants' abilities for parsing the ongoing activity in their surroundings into distinct and meaningful parts. However, the individuation of actions is a fundamental ability, as explicated in a four-part introduction. Based on a review of general principles of individuation across multiple ontological domains, three possible mechanisms for action individuation in infants are identified and tested. The results of a series of studies show some important limitations in infants' abilities to parse actions from continuous motion. Although infants can perceptually discriminate different types of actions (such as jumps and falls) performed by a puppet, and can individuate and enumerate sequences of such actions when the acts are separated by brief motionless pauses, their ability to individuate actions embedded within a continuous strewn of motion is limited: Neither repeating cycles in the action sequences nor marked differences in extent of motion are sufficient cues. The results instead suggest that tangent discontinuities in the path of motion are an important cue to infants' ability to parse actions from on-going motion. Implications for infants' conceptual structure for actions, and additional potential mechanisms of action individuation, are also discussed.

Semantic plausibility has been shown to affect a number of sentence processing tasks, including reading, sentence matching, and RSVP (Rapid Serial Visual Presentation). In this thesis it is argued that anomaly, and not plausibility is the critical variable. Unfortunately the distinction between anomaly and implausibility has not traditionally been a clear one, and definitions can vary depending on the semantic theory that is being adopted. The experiments reported in this thesis are aimed at finding a clearly definable distinction between anomalous and implausible sentences, and to show a reliable empirical consequence of the distinction. The major emphasis is placed on the sentence matching task, which has been claimed to be sensitive to a very specific level of syntactic processing, and to be unaffected by extraneous variables. Experiments 1 and 2, however, demonstrate the very clear effect of a semantic manipulation, on the sentence matching task. Experiments 3, 4, and 5 were designed to refine our understanding of the aspect of the manipulation that was critical. Finally in experiment 6, it is argued that the violation of basic conceptual categories which have been argued to organize lexical/conceptual structure, is the only circumstance under which the effect of meaning is seen in the sentence matching task, and possibly the reading task. Experiment 7 shows that the same conclusion is true for the RSVP task. These conclusions are considered in the light of Jackendoff's (1983) and Pustejovsy's (1990) theory of semantics. It is further argued that a close symbiotic relationship can exist between these formal theories, and the empirical findings.

The main focus of this dissertation is the processing of agreement between a subject and predicative adjective in Spanish. The basic methodology often employed in experiments in the literature on subject-verb number agreement was used-participants saw an adjective and a sentence preamble consisting of a subject head noun and a prepositional modifier containing a second noun (attractor), and they had to complete the sentence by adding a verb and using the given adjective. Agreement errors in the gender and number of the adjective and in the number of the verb were analyzed. In the first experiment, the possible differences between two types of gender were studied. Spanish nouns can be divided into two types according to whether their gender has semantic import. Most nouns referring to animate beings possess gender specification at two levels-semantic and grammatical, while the rest possess gender only at a grammatical level, and this difference could be reflected in the agreement process. The results point to the fact that gender agreement is indeed sensitive to gender type. The second experiment investigated the relationship between gender and number agreement with the predicative adjective, and of number agreement between subject and verb, and subject and predicative adjective. The results indicate that processing of gender agreement is independent of processing of number agreement. They also indicate that the computation of number agreement between a single source (subject head) and different targets (verb and predicative adjective) is a single process, that is, that agreement with one of the targets is dependent on agreement with the other. In the third experiment, the effect of morphophonology on gender agreement was tested by manipulating the regularity of the nouns in the sentence preambles. The pattern of errors implies that the morphological properties of nouns with semantic gender does affect agreement, but not so the phonological properties of nouns with purely grammatical gender. Overall, the results of the three experiments are more compatible with a modular model of language production, as well as with a feature copying account of agreement implementation.

Traditional epistemology has devoted much attention to the distinctions between perception and inference and between basic and non-basic beliefs. Hot, I develop a different and more general distinction, between what I call "privileged" and "nonprivileged" beliefs; privileged beliefs are justifiable by means of an otherwise substandard argument while nonprivileged beliefs require support by a generally adequate argument for their justification I argue that even coherentists are tacitly committed to this distinction (although they may deny the existence of basic beliefs) and that one of the chief problems for simple reliabilist theories is that they imply that all beliefs are privileged. Any adequate epistemology has to count some beliefs as privileged and some as nonprivileged, and I suggest a way to modify reliabilist theories to accommodate this result. The privileged/nonprivileged belief distinction suggests a framework theory about the structure of epistemic justification, a theory which improves on foundationalism, coherentism, and reliabilism in certain respects. Yet it raises the question of which beliefs are privileged and which are nonprivileged. I argue that whether or not a belief is privileged is determined by the etiology of that belief, and in particular, by the intrinsic nature and the etiology of the psychological faculty that produced that belief. A belief, therefore, is privileged if and only if it is the output of a certain kind of cognitive faculty, or system. Consequently, the beliefs produced by these faculties are such that it is possible to be justified in holding them even in the absence of a generally adequate argument. This does not mean that all the outputs of all such faculties are justified, for such beliefs might still require (and lack) inferential support or be subject to non-inferential requirements, like reliability And of course, all such beliefs are potentially subject to defeat from other justified beliefs. The kind of cognitive faculties I have in mind includes, but is not restricted to, "modules", in Jerry Fodor's sense. The etiological, faculty-oriented view defended hat argues for distinctive, versions of externalism and naturalism in epistemology and holds some promise of illuminating certain traditional epistemological problems.

The current theoretical formulation of bilingual memory (The Process View of Memory) assumes that all bilinguals are the same and thus attributes cross-language memory transfer effects to the processing components of memory tasks alone. However, the present study found that only early, but not late, bilinguals exhibited significant cross-language transfer effects in the implicit memory word fragment completion task under separate encoding conditions that involved perceptual, conceptual, and integrative processing (i.e., reading, imaging, and sentence processing). Results are taken to suggest that early and late bilinguals adopt different information processing strategies at encoding and retrieval, and question the notion of task processing demands as the only or main determinant of bilingual memory transfer. The present findings help explain the pattern of inconsistent bilingual transfer effects that have emerged in research studies conducted under the Process View of Memory and support the move toward the adoption of a bilingual memory paradigm in which cross-language transfer is explained in terms of both subject and task dimensions.