29.01.2013 Views

Abstracts 2005 - The Psychonomic Society

Abstracts 2005 - The Psychonomic Society

Abstracts 2005 - The Psychonomic Society

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Papers 307–313 Sunday Morning<br />

may have a promotion focus (a sensitivity to potential gains in the environment)<br />

or a prevention focus (a sensitivity to potential losses).<br />

Regulatory fit suggests that people perform best when their regulatory<br />

focus matches the payoffs in the environment. We generated a<br />

category structure in which a unidimensional rule led to good but subcriterial<br />

performance. <strong>The</strong> learning criterion could be achieved, however,<br />

using a complex conjunctive rule. People were best able to find<br />

the conjunctive rule when there was a regulatory fit. If there were rewards<br />

on every trial, people with a promotion focus learned better<br />

than people with a prevention focus. If there were punishments on<br />

every trial, people with a prevention focus learned better than people<br />

with a promotion focus. <strong>The</strong>se findings suggest that regulatory fit<br />

leads to flexible cognitive processing.<br />

11:00–11:15 (307)<br />

Varying Abstraction in Categorization. GERT STORMS & WOLF<br />

VANPAEMEL, University of Leuven—A model is proposed that elegantly<br />

unifies the traditional exemplar and prototype models. <strong>The</strong>se<br />

two models are extreme cases of the proposed varying abstraction<br />

model. This unifying model further makes room for many new intermediate<br />

pseudoexemplar models. Analysis of empirical data, using<br />

both artificial stimuli and natural concepts, showed evidence in favor<br />

of intermediate pseudoexemplar models as compared with the classical<br />

prototype and exemplar models. A further advantage of the varying<br />

abstraction model is that because of its pseudoexemplar models,<br />

it allows for sensitive modeling. For example, the model is highly useful<br />

for investigating in full detail recent findings that category representation<br />

may change during the learning process. <strong>The</strong> model can also<br />

capture possible varying categorization strategies depending on the<br />

complexity of the categories under study.<br />

11:20–11:35 (308)<br />

Does Response Scaling Cause the Generalized Context Model to<br />

Mimic a Prototype Model? DANIEL J. NAVARRO, University of Adelaide,<br />

& JAY I. MYUNG & MARK A. PITT, Ohio State University (read<br />

by Jay I. Myung)—Exemplar and prototype accounts of categorization<br />

phenomena differ primarily in terms of the manner in which category<br />

information is represented. However, the models also make auxiliary<br />

assumptions that are instantiated as model parameters. Smith and Minda<br />

(1998) argue that the response scaling parameter in the exemplarbased<br />

generalized context model (GCM) makes the model unnecessarily<br />

complex and allows it to mimic prototype representations. We<br />

estimate the complexity of GCM with and without the parameter, as<br />

well as that of a prototype model. We then go on to assess the extent<br />

to which the models mimic each other’s behavior. <strong>The</strong> parameter does<br />

increase the complexity of the model, but this complexity only allows<br />

partial mimicry. Furthermore, if we adopt minimum description<br />

length as the measure of model performance, the models become<br />

highly discriminable.<br />

11:40–11:55 (309)<br />

Effects of Noun Modification on Plausibility of Attribute Information.<br />

JAMES A. HAMPTON, City University, London, & MARTIN<br />

JONSSON, Lund University—Inspired by a recent paper by Connolly<br />

et al., we investigated whether attributes judged to be generally true<br />

of a noun class will also be considered true of an orthogonal subset<br />

of that class. For example, given that ravens are black, how likely is<br />

it that young jungle ravens are black? Experiment 1 replicated their<br />

result. Modification, particularly with atypical modifiers, reduced the<br />

judged truth of typical attributes. Experiment 2 required forced choice<br />

decisions of likelihood between modified and unmodified subject<br />

nouns and asked for additional justifications. Experiment 3 showed<br />

that the reduction in judged likelihood was correlated with the mutability<br />

of the predicate for the subject noun. Finally, we demonstrate a<br />

new version of the conjunction fallacy. People judged a universally<br />

quantified statement (All ravens are black) to be more likely to be true<br />

than the same statement with a modified subject noun (All jungle<br />

ravens are black).<br />

48<br />

Visual Word Recognition<br />

Grand Ballroom West, Sunday Morning, 10:00–12:00<br />

Chaired by Wayne S. Murray, University of Dundee<br />

10:00–10:15 (310)<br />

<strong>The</strong> Form of the Word Frequency Effect. WAYNE S. MURRAY, University<br />

of Dundee, & KENNETH I. FORSTER, University of Arizona—<br />

Murray and Forster (2004) made strong claims about the precise form<br />

of the effect of word frequency on recognition time. It was argued that<br />

the underlying function directly reflects the relationship between access<br />

time and the rank order of words in a frequency-ordered search<br />

or verification set. Across three experiments, a rank function produced<br />

an account that was numerically superior to the more “traditional”<br />

description of this function as logarithmic, but no clear statistical<br />

superiority was demonstrated. In this paper, we consider word<br />

recognition data derived from a number of sources and show that although<br />

a log function provides a reasonable approximation across the<br />

mid-frequency range, it is clearly inadequate at the frequency extremes,<br />

whereas rank produces a good fit throughout the entire frequency<br />

range. We discuss both the practical and theoretical consequences<br />

of this finding.<br />

10:20–10:35 (311)<br />

1984: Reconsidering the Absence of Frequency Attenuation Effect<br />

With Masked Repetition. SACHIKO KINOSHITA, Macquarie Centre<br />

for Cognitive Science—<strong>The</strong> frequency attenuation effect refers to the<br />

finding that the magnitude of frequency effect in the lexical decision<br />

task is reduced for repeated words. Forster and Davis (1984) reported<br />

that the frequency attenuation effect is not found with masked repetition<br />

but that instead effects of frequency and masked repetition are additive.<br />

This pattern has been replicated many times (although see Masson<br />

& Bodner, 2003, for contradictory data), and has been taken as a<br />

benchmark finding for accounts of masked repetition priming. I report<br />

contrary data showing that a frequency attenuation effect is found<br />

with some low-frequency words and that this is not due to methodological<br />

artifacts, such as greater power or stronger manipulation of<br />

frequency. Implications of the results for the nature of lexical representation<br />

and the lexical decision task will be discussed.<br />

10:40–10:55 (312)<br />

Visual Word Recognition: Is a Hybrid Connectionist Framework<br />

Needed? JENNIFER A. STOLZ, NASEEM AL-AIDROOS, & DEREK<br />

BESNER, University of Waterloo—Interactive activation is currently<br />

the dominant conceptual framework for language processing. That<br />

said, such a framework encounters difficulty in explaining the combination<br />

of several empirical observations in the context of the lexical<br />

decision task (an interaction between the effects of stimulus quality<br />

and semantic priming when the relatedness proportion is .5,<br />

additive effects of these same two factors when the relatedness proportion<br />

is .25, and additive effects of stimulus quality and word frequency).<br />

We report a series of simulations from a hybrid model that<br />

combines interactive activation across later stages with an earlier<br />

stage that is thresholded. This model produces all of the phenomena<br />

noted above, as well as many others.<br />

11:00–11:15 (313)<br />

Early Semantic Effects in Visual Word Recognition. KENNETH I.<br />

FORSTER, University of Arizona—Several phenomena have now<br />

emerged suggesting that in semantic categorization experiments, the<br />

semantic properties of the category fundamentally alter very early<br />

processes in visual word recognition. For example, masked semantic<br />

priming with a short-duration prime occurs reliably in semantic categorization<br />

but not in lexical decision. Similarly, masked cross-language<br />

translation priming from the second language to the first is obtained<br />

only in semantic categorization. One possibility is that these effects<br />

might be explained as congruence effects. However, there is at least<br />

one effect that cannot involve congruence—namely, the greater sen-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!