LECTURE 2.7 – The Meaning of Meaning
Semantic Memory is a general knowledge of concepts and it is part of our long term memory
Collins & Quillian developed a hierarchical model of semantic memory. In this model concepts
are represented by nodes, and these nodes are connected to each other through links. So for
example bird would be linking to animal and features such as can fly, has feathers, chirps, etc.
would all be linked to bird.
One way to test this model is through semantic priming. Priming tasks have revealed that when
the prime is semantically linked to the target word, our lexical access is faster.
The term stimulus onset asynchrony (SOA) refers to the time lag between the onset of the end
of the prime and the onset of the stimulus. The SOA becomes important when considering
So priming semantically related words increases facilitation and improves reaction times in
lexical decision tasks. What if the words are unrelated? When the prime of a word is
semantically unrelated there is an inhibition effect. This inhibition effect only occurs however
when the SOA is about 500 msecs.
These effects are predicted by the Collins & Quillian Semantic Network Model. When a prime is
semantically related, it should activate a concept very near it in your semantic network. The
opposite is true for semantically unrelated words, they should lead you into the wrong are of
your semantic network.
Why is it that the inhibition only occurs at SOA’s of 500 msecs and greater? We speculate it’s
because that’s how long it takes for people to form conscious expectations.
In priming experiments we also see frequency and typicality effects. For example a bug usually
refers to an insect as opposed to recording devices. Therefore when primed with something like
ant reactions times will be faster than if the prime were microphone. A robin is a more typical
member of the bird category than a penguin is and therefore bird will prime better for the word
robin than it will for the word penguin.
Neely conducted an interesting study on priming. He trained participants that the word
BODYPART will always precede a word that refers to part of a building.
o i.e. BODYPART DOOR
What he found was the BODYPART primed for DOOR faster than it did for words like HEAD.
These findings suggest that our semantic memory is very flexible.
McRae & Cree Model of Semantic Memory. This model is very modular in the sense that
information flows in one direction only and is carried out by a special processor. Essentially the
model consists of two parts; it has an input layer which picks up phonological forms and it has an