Three experiments investigated the ways memories change over repeated tests. A tradition of previous work started by Ballard (1913) revealed net improvement (or hypermnesia) over tests, while Bartlett (1932) showed intertest forgetting. Differences between the paradigms include the type of material learned (lists of items vs. prose) and the delay interval between recall tests (several days for Bartlett and up to one day for Ballard). Both factors were examined in three experiments. In all experiments, hypermnesia was obtained across repeated memory tests when tests occurred within a single, experimental session. However, intertest forgetting resulted after a one-week delay, replicating Bartlett. Performance on repeated tests was discovered to be a function of the number of prior tests and the delay between tests, while response criterion (lenient or strict) and the nature of the learned material (pictures or prose passages) were not important.

Research has shown that generating words in a study phase (e.g. "shoe" from foot-s$\sb{--}$e) usually yields better memory performance compared to reading words (e.g. foot-shoe), and this advantage for generated words relative to read words is called the generation effect. One particularly interesting version of the generation effect was studied by Jacoby (1978). In two conditions of his study, subjects generated a word once (e.g. foot-s$\sb{--}$e) or generated a word immediately after reading it (e.g. foot-s$\sb{--}$e preceded by foot-shoe). Jacoby (1978) found that generating a word once produced better cued recall (e.g. foot-? ? ? ?) than reading a word, then generating it. One purpose of the present experiments was to investigate if Jacoby's (1978) results would generalize to other memory tests. The experiments reported manipulated the number of study presentations, the modality of study presentation, the difficulty of generating a word, and the match between perceptual cues (e.g. s$\sb{--}$e) presented at study and test, and examined performance on four tests (associative cued recall, recognition, word fragment cued recall, and word fragment completion). Jacoby's (1978) results were replicated for associative cued recall, but other patterns of data were obtained on the other tests. A new finding for a word fragment cued recall test occurred based on the match of perceptual cues between study and test. Presenting the same word fragment (e.g. s$\sb{--}$e) at study and test resulted in better memory performance than did presenting different word fragments at study and test cue (e.g. s${--}$e at study but s$\sb-$o$\sb-$ at test). Dissociations between word fragment cued recall and the word fragment completion ruled out the possibility that generation effects obtained for the word fragment completion test were due to subjects using intentional retrieval strategies. Most of the results were explainable with the transfer appropriate processing framework, which states that the greater the overlap in mental processes engaged at study and test, the better test performance will be.

Modern automotive instrument panels are often equipped with numeric readout and digital bargraph displays instead of traditional analog displays. An informal survey of 1990 model cars revealed wide disparity in automotive displays. Three major classes of displays were in use: (1) traditional analog, (2) binary indicators, and (3) readout displays. The present work reports results from four experiments of display monitorability. Twelve computer-simulated displays were designed for evaluation by dual-task methodology. Subjects were required to maintain performance on a demanding tracking task while monitoring configurations of four-displays for "critical readings." Subjects' latencies to respond to such readings and their tracking error scores were collected. Experiment 1 examined displays in homogeneous clusters. Orientation, configuration and class were studied. Experiment 2 compared displays in heterogeneous configurations. Experiment 3 studied the addition of color to the displays to facilitate the detection of a critical reading. Static color (i.e., a "red zone") was added to the analog displays while dynamic color was added to the bargraph displays. Finally, Experiment 4 examined the effect of check-readable layouts for analog and bargraph displays. Older subjects were slower to respond to all displays and exhibited poorer tracking performance. They also had more variability than younger subjects on both of these measures. This result was likely due to age-related slowing rather than changes in cognitive processes. No class of displays studied emerged as superior, but the binary indicator (a modified warning lamp) generated the fastest responses. Although no evidence was found for effects of orientation, configuration or check-readability, support was found for color facilitation. Male subjects tended to respond faster when static color was present, and age-related facilitation was found for dynamic color. The addition of dynamic color to displays monitored by subjects in the middle and older age groups approximately compensated for age-related RT decrements. It is likely that displays which incorporate dynamic color elements will be most useful for presenting rate information as well as calling attention to off-normal readings. Pending further study, it is recommended that designers use new display technology with caution.

An exploratory study was conducted with users of a command-based system, MS-DOS, and the Graphical User Interfaces (GUI) of the Apple Macintosh. It was hypothesized that command-based interfaces and GUIs differ in the ease with which they afford attainment of particular concepts, and that in general the concepts investigated would be more readily attained by Macintosh users. The study attempted to assay subjects' knowledge of particular computer-related concepts, their ability to perform related operations, and the organization of that knowledge. Contrary to some theory (e.g., Tennyson and Cocchiarelli, 1986), a double dissociation between verbalizable conceptual knowledge and performance was observed. Results of the study suggested differences in the support provided by DOS and the Macintosh interface for development of knowledge of particular concepts and procedures, underscoring the potential value of understanding the detailed effects of particular interfaces, and classes of interface, upon user knowledge.

The prior presentation of a word enhances the probability that the word will be identified from a degraded cue. This enhancement, or priming, may occur without recollection of the prior presentation. In investigations of the effect of delay on nonrecollective memory, priming of word completion may occur at long delays after the initial word presentation (e.g., HORIZON) when the degraded cue contains scattered letters (e.g., R Z) but not when the degraded cue contains the initial letters (e.g., HOR$\underline{\qquad}).$ These particular cues are known as fragments and stems respectively. Priming of fragment completion may occur weeks after the study presentation whereas priming of stem completion may disappear within hours. The present research investigated this apparent discrepancy between the duration of priming effect for the two types of cues by examining completion of both within the same experiment. Particularly, it tested the hypothesis that the decline in priming of completion is greater when cues have many possible completions. Subjects completed fragments that have 1 completion or fragments and stems that have more than 7 completions. Priming was measured over a 48-hour delay interval in each of 4 experiments. Priming of stem completion declined at a greater rate than priming of fragment completion over the 48-hour interval, thus replicating within a single experiment the previous between-experiment findings. The number-of-completions hypothesis was not rejected in Experiments 1 and 3 because the decline in priming of multiple fragment completion did not differ from the decline in priming of stem or unique fragment completion. However, in Experiment 2 the decline in priming of multiple fragment completion differed from the decline in priming of stem completion but not from the decline in priming of unique fragment completion. Thus, the number-of-completions hypothesis may not be a viable explanation of the differential effect of delay on priming of fragment and stem completion. When the same words served in both the unique fragment and stem conditions, the effect of delay on priming of completion was similar for both conditions (Experiment 4).

People rely extensively on inference as a source of information, and sometimes they confuse inference with observation. Specifically, inferred information is sometimes mistaken for observed information. Such confusion of inference with observation can be problematic, especially if the inferred information is erroneous. One factor that might affect the probability of mistaking inference for observation is the degree of consistency between inferred information and subsequently encountered information. The present research was designed to test this possibility. In three experiments subjects made inferences on the basis of presented information, and then were given additional information that was varied in consistency with the information they had inferred. Finally, subjects were tested for accuracy in discriminating the source (inference vs. observation) of the inferred information. As expected, accuracy was lower when subsequently presented information was relatively consistent with inferred information than when it was relatively inconsistent with inferred information. This effect did not vary with delay between making an inference and attempting to discriminate information source. It is concluded that consistency of inferred information with subsequently encountered information can affect the probability of mistaking inference for observation, with the probability of error increasing as consistency increases.

This project was designed to examine the nature of recollective experience. Gardiner (1988) showed dissociations are obtained within an explicit memory test as a function of several independent variables when subjects classify recognized items into two types of recollection, one termed "Remember" (those items that subjects consciously recollect from the study list) and the other termed "Know" (those items that subjects recognize on some other basis, in the absence of conscious recollection). Gardiner and his colleagues concluded from their results that "Remember" responses are sensitive to conceptual and "Know" responses to perceptual processing. The present investigation tested this conclusion using the recognition memory task. In Experiment 1 subjects produced semantic associates to some and rhyme associates to other study items. Levels-of-processing effects were observed for recognition and "Remember" judgments. For "Know" judgments, this effect was reversed. A modality match (visual - visual) between study and test events had no effect relative to a modality mismatch (auditory - visual) condition. In Experiment 2, subjects studied pictures and words. The picture superiority effect was obtained for recognition and "Remember" judgments and was reversed for "Know" judgments. In Experiment 3, in the recognition phase, half the studied and nonstudied words were preceded by a brief (50 ms) and masked repetition, and the other items were preceded by masked presentations of unrelated words. Masked repetition enhanced recognition and "Know" responses, but did not affect "Remember" responses. In Experiment 4, subjects made confidence judgments ("Sure"/"Not Sure") to primed or unprimed recognized items. The pattern of priming effects for "Sure"/"Not Sure" responses and "Remember"/"Know" responses (Experiment 3) were dissimilar. In Experiment 5, the primes in the recognition test list were semantically related or unrelated to the test words and were presented longer than repeated primes (SOA = 250 ms). The semantic priming effect was observed only for nonstudied items and was distributed equivalently between "Remember" and "Know" judgments. These results: (1) indicate that "Remember" and "Know" responses are sensitive to conceptual and perceptual factors respectively, (2) provide support for the two-factor theories of recognition memory, and (3) show that "Remember" responses are a "purer" measure of conscious recollection than standard explicit memory tests.

People are often biased in their perception of randomness in that they tend to see patterns in random distributions. This is a serious problem because the accurate perception of randomness can be important in decision-making. This research presents three experiments investigating (a) the generality of this bias in randomness perception and (b) the trainability and transfer of the ability to judge randomness. The three experiments replicated the finding that people tend to judge random sequences as systematic. However, no evidence was obtained that the same bias occurs in judging lottery outcomes, thus calling into question the generality of the bias. No evidence was found for reducing people's perception bias with a short period of training. No evidence was found for far transfer though near transfer did occur.

Four experiments are presented which explore the effects of warning layout (spatial structure) and organization (semantic structure) on the ability to recall and the tendency to read warning information. Warnings were either presented in paragraph layout or in a meaningfully indented "outline layout." Warning information was organized by hazard, type of warning statement, or randomly. Experiment 1 results indicate that layout and organization reliably influenced perceptions of warning eye appeal, ease of processing, and effectiveness. However, a test of recall in Experiment 2 did not reflect the effects of these variables. In Experiments 3 and 4 warnings were presented to subjects in an incidental warning paradigm. The results of these experiments indicate that warnings in outline layout elicited less variability in reading rates, were more likely to be read, and were more likely to elicit warning compliance. Implications for warning design and future research are discussed.

Three experiments explored the extent to which word fragment completion and word stem completion could be "primed," or facilitated, by prior study of the words. In Experiment 1, the manner in which the words were studied beforehand had little effect on priming of either kind of completion, but delaying the test reduced the amount of priming. More importantly, priming of fragment completion decreased over the delay to a greater degree than priming of stem completion. In Experiment 2, this interaction was not replicated when both fragments and stems were constructed without controlling the number of possible completions and were placed in the same test. In a third experiment, the number of response alternatives did not affect priming of stem completion. It was concluded that differences in the amount of priming of fragment completion and stem completion cannot be easily explained by the number of response alternatives or by the task difficulty.