Memory conjunction error is a common phenomenon that occurs when we incorrectly combine parts of previously experienced memories to create an entirely new memory. For example, such an error has occurred when a person remembers seeing the word toothache after viewing the words toothpick and earache instead. Two theories have been proposed in the literature to account for the mechanisms underlying such errors. In one account, memory conjunction errors occur because features stored in episodic memory are incorrectly conjoined. In another account, memory conjunction errors occur simply because conjunction lures seems familiar. In an attempt to distinguish the two theories, the current research focuses on the differences between retrieval mechanisms. Experiment 1 introduces and examines an important factor, the accessibility of episodic features. Experiment 2 further confirms that feature accessibility plays an important role in the occurrence of feature errors. Experiment 3 investigates the retrieval dynamics of the two types of episodic features. The current data pose major problems for the binding theory's claim about a feature binding process occurring at retrieval. Taken as a whole, the current data also show that unified events can be represented differently in episodic memory depending on the nature of the associated features.

This work investigated the ways in which the properties of a concept are activated when that concept is accessed. There has been considerable debate about how closely property information is tied to concepts and under what conditions it is available (e.g., Margolis and Lawrence, 1999). If property information is automatically activated, it should be detectable in both frequency estimation and speeded response tasks. According to Barsalou and Ross' automaticity hypothesis (1986, p.117), “…people become sensitive to the frequency of non-presented information through automatic processing of presented items by well-established memory structures.” On this account, if a list of concepts is presented, participants may be sensitive to the frequency of their properties. Therefore, after studying a list of items, participants should be able to estimate the number of items that were “red” or “sweet” without recalling individual items. Naturally, some properties are more important to a concept than others and are more likely to be activated. Sloman, Love, and Ahn (1998) developed a taxonomy of conceptual properties. Using ratings obtained in a variety of tasks, they performed a factor analysis that revealed three factors: centrality, salience, and diagnosticity. In these studies, I manipulated centrality and salience to appraise their relative importance for the activation of properties. Barsalou's (1982) work on context-independent and context-dependent properties asserts that the activation of properties may be automatic or strategic, depending on the property type. In both a frequency estimation task and a sentence-word priming task, I manipulated context to evaluate whether central or salient properties are context-dependent. In the sentence-word priming task, I was also able to assess degrees of context dependency. I found that: (1) people demonstrated frequency sensitivity to both central and salient properties but were more sensitive to central properties; (2) central properties appear to be activated faster than salient properties as indicated by the slopes in the frequency estimation task and reaction times in the sentence-word priming task; (3) the activation of both central and salient properties appear to be context-dependent or situated (Barsalou, 2000), with the activation of central properties being moderately context-dependent and that of salient properties, highly context-dependent.

The present research investigates the relationship between spoken language and reading processes by using several experimental techniques to examine the nature of the phonological representations used during silent reading. Experiments 1 through 4 measured eye movements during sentence reading and lexical decision using a parafoveal preview paradigm. In Experiment 5, brain electrical potentials were recorded in a four-field masked priming paradigm during passive reading of single words. Experiments 1 and 2 asked whether the phonological representations used by skilled readers in lexical access are minimal and contain only consonant information, or whether they include phonological vowel information as well. Experiments 3, 4, and 5 examined whether the phonological representations used in lexical access contain prosodic information about syllables as well as phoneme information. In combination, these experiments demonstrated that skilled readers are sensitive to vowel and prosody information presented in parafoveal previews and masked foveal primes. This suggests that readers routinely activate elaborate, speech-like phonological representations early in word recognition during silent reading. The phonological hub theory of silent reading is proposed to account for this finding and situate orthographic and phonological processes in the context of natural silent reading.

Compound words are morphologically complex words that are composed of two lexemes (e.g. farmhouse, blackboard). By investigating how compound words are processed, insight can be gained into the organization of the mental lexicon. Five experiments are reported which explored how English compound words are processed. In each experiment, compound words were embedded into sentences and readers' eye movements were recorded as they read these sentences. Several reading time measures were analyzed to investigate the time-course of compound word recognition. Representations of the compound words were examined by manipulating the frequency of the beginning lexeme in the compound (Experiments I1A, 1B, 3A), the frequency of the ending lexeme (Experiment 3B), or the familiarity of the whole compound word (Experiment 2). Experiments 1A and 1B demonstrated that compound word length does not consistently modulate whether beginning lexeme frequency effects are observed for English compound words. However, there are larger effects for overall compound familiarity on most reading time measures for long compound words compared to short compound words (Experiment 2). Experiments 3A and 3B manipulated the predictability of the sentence context with respect to compound words. Having a highly predictive sentence context significantly reduced early effects of beginning lexeme frequency (Experiment 3A), but not later effects of beginning lexeme frequency (Experiment 3A) or ending lexeme frequency (Experiment 3B). Also, compound words that contained a high frequency lexeme were read faster than length and frequency matched monomorphemic words. Correlations and multiple regression analyses suggested that the size of a compound word's morphological family and the number of higher frequency morphological family members significantly affect compound word reading time. The results are discussed in terms of a proposed framework for English compound word recognition where morphological representations are hierarchically organized within the lexicon.

Many studies show that bottom-up and top-down information interactively control attentional deployment. This study explores how these two factors are integrated when controlling attention and eye movements. In this study, attention was affected by top-down factors (informative location cues, strategies) and by a bottom-up factor (an orientation singleton), with the intensity of each manipulated systematically. Results of Experiment 1 showed that one top-down factor (previous knowledge about the target location) can control attention independently from the bottom-up factor. This result raises some difficulties for those models that predict competition between top-down and bottom-up factors to control attention. The study implies separation between the pathways for top-down attention control and bottom-up attention control. This pattern is consistent with recent neuroscience findings, which show that different brain regions are involved in top-down and bottom-up attention control. Strategies were manipulated between Experiment 1 and Experiment 2. In Experiment 1, participants were encouraged to employ a singleton detection mode; in Experiment 2, participants could not employ a singleton detection mode. Results showed that the singleton could capture attention in Experiment 1; however, only the most salient singletons could capture attention in Experiment 2. This result is only partially consistent with the idea that a singleton could capture attention only when participants employ a singleton detection mode. Singletons with some specific properties could apparently capture attention even when participants could not employ singleton detection mode. Experiment 3 showed that eye movements are also controlled by the interaction of the top-down factors and the bottom-up factors. First saccades went to the singleton location more often if participants employed singleton detection mode than if they used a different strategy. First saccades also went to the singleton location more often when the informativeness of the cue decreased, and when the orientation of the singleton increased. Eye movement control differed from covert attention control in that it showed competition between top-down and bottom-up factors. This difference in overt and covert attentional control probably arises because covert attention can be simultaneously split across multiple locations, while the eyes can only be directed to a single location at any one time.

Thesis (M.S.)--University of Michigan, 1981.

Thesis (M.S.)--University of Michigan, 1981.

published_or_final_version / Clinical Psychology / Master / Master of Social Sciences

published_or_final_version / Clinical Psychology / Doctoral / Doctor of Psychology

Value convergence refers to the empirical finding that in successful psychotherapy patients adopt their therapist's values. This study examines the relationship between therapy outcome and the following predictor variables: The initial similarity of patient and therapist values; patient sensitivity to therapist values; value convergence, changes in dysfunctional beliefs, and patient's ability to predict therapist's values. Previous attempts to define values are examined as well as conceptualizations of the relationship of values to psychotherapy. The empirical research relating to value change and psychotherapy is reviewed. A growing body of literature has largely confirmed the value convergence phenomenon. Methodological weaknesses in this literature are discussed. The related research area in cognitive therapy concerning the relationship of belief changes and depression is also reviewed, followed by a summary of cognitive therapy (CT) theory and practice. A total of 29 depressed older adults were randomly assigned to group CT or to a medication (alprazdam)/supportive therapy condition. The Rokeach Value Survey (RVS) assessed value similarity, value convergence and subject predictions of therapist values. A scale developed here, the Treatment Sensitivity Survey (TSS), assessed sensitivity to therapist values. The Cognitive Error Questionnaire (CEQ) measured changes in dysfunctional cognitions. Of these variables only sensitivity to therapist values significantly predicted improvement as measured by the Beck Depression Inventory. Initial values similarity also predicted value convergence. Subjects in the CT condition evidenced greater value convergence and made more accurate predictions of therapist values.