Recent Conference Talks
SLAT RoundTable Talk, Feb 2004
Click here for presentation.
Semantic Processing of Unconsciously Perceived Words
Recently, there has been some debate about whether truly unconsciously perceived words receive a semantic interpretation. Evidence from masked semantic priming experiments is ambiguous, since priming in lexical decision is obtained only with relatively long SOAs (60 ms), suggesting the possible involvement of awareness. Congruence effects in semantic categorization tasks (slower responses when the prime belongs to a different category than the target) are also problematic in that the effects apparently require pretraining on the words that will later be used as primes. Several experiments will be reported that show unambiguously that semantic congruence effects can be obtained at short SOAs (55 ms) without pretraining. However, these effects depend on the type of category involved.
Mental Lexicon Conference, 2002
Lexical Processing Anomalies in Task Comparisons
Analyzing the similarities and differences between the effects of different variables in semantic categorization (SC) and lexical decision (LD) tasks can provide valuable clues to the nature of the lexical processing system. Several effects standardly obtained with LD appear to be absent in SC tasks, or only present with specific types of categories. Among these are ambiguity effects, frequency effects, congruence effects, and neighborhood effects. However, phenomena such as masked repetition priming and form priming appear to be identical in both tasks. One recent case that is particularly puzzling is the finding of a masked repetition effect in SC for nonexemplars despite a complete absence of any frequency effect. It is argued that this effect can be understood as the natural consequence of decision processes.
Knowing What Is Not: Frequency and Masked Priming Effects in Semantic
Balota and Chumbley (1984) reported an absence of a frequency effect for
non-exemplars in a semantic categorization task, triggering a debate about
the role of frequency in lexical access. Subsequent research has confirmed
the presence of a frequency effect, at least for large categories (e.g.,
"animals"). However, for small categories (e.g., "months"), it appears
they were correct, raising the question of how we can know that a word
is not a month without accessing its lexical entry (or activating its
semantic properties). This puzzle is further complicated by the fact that
there is a clear masked priming effect for non-exemplars in such a task,
which is usually taken to indicate that lexical access has occurred. A
solution to the puzzle is offered which involves postulating a category
Lexical Interactions in Dense Neighborhoods
It has been suggested that the absence of masked form-priming (e.g.,
race-FACE) in a dense neighborhood occurs only when the prime and
the target share a neighbor, e.g., pace is a neighbor of both race
and face, but fact is not (Van Heuven, Dijkstra, Grainger
& Schriefers (PB&R 2000). Several experiments are reported using low-,
medium-, and high-density targets. There was no indication that the amount
of form-priming was influenced by the presence of shared neighbors. Furthermore,
no priming was found for high-density targets that did not share a neighbor.
Masked Priming Effects in Lexical Decision and Semantic Categorization
Important clues to underlying processes can be revealed by examining how
priming effects vary across tasks. In this study, the relative strengths
of identity priming (music-MUSIC) and form priming (nusic-MUSIC) are measured
at each of three different prime-target SOAs (35 ms, 46 ms, and 57 ms)
for three tasks: lexical decision, semantic categorization with a natural
category, and semantic categorization with an ad hoc category. The questions
of interest are: (a) is identity priming a linear function of SOA for
all tasks? (b) Is form priming stronger in a task emphasizing orthography
(i.e., lexical decision)? (c) Is priming in semantic categorization the
same for exemplars and nonexemplars? (d) Does priming depend on the nature
of the category?