Cognitive Semantics : Meaning and Cognition
Cognitive Semantics : Meaning and Cognition
Cognitive Semantics : Meaning and Cognition
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
174 JORDAN ZLATEV<br />
which need to be combined recursively before there can be anything resembling<br />
concepts. The result has been something of a debate between proponents of<br />
“pure connectionist models” <strong>and</strong> “hybrid connectionist / symbolic models” of<br />
the grounding of meaning. However, linking embodiment <strong>and</strong> connectionism<br />
directly is likely to lead to confusion. For one thing, connectionism covers a<br />
whole spectrum of models that vary on properties such as pre-structuring <strong>and</strong><br />
biological realism. Secondly, <strong>and</strong> more importantly, connectionism is a new<br />
kind of technology <strong>and</strong> thus provides a new kind of metaphor for the mind; but<br />
it is not a theory about the mind <strong>and</strong> can be used even by conflicting theoretical<br />
st<strong>and</strong>points.<br />
Therefore, I will take a step back <strong>and</strong> characterize independently from<br />
neural nets one particular approach to linguistic meaning <strong>and</strong> embodiment.<br />
For reasons that should become clear in Section 2, I will be referring to this<br />
approach as situated embodied semantics. In Section 3 I briefly describe<br />
Regier’s system, mentioned above, <strong>and</strong> use it as an illustration of (some<br />
aspects of) the situated embodied approach. Section 4 discusses how “the<br />
creativity of language”, i.e. the ability to utter <strong>and</strong> comprehend novel expressions,<br />
could be explained from this perspective, <strong>and</strong> in Section 5 I present<br />
what I think is the main reason for the shortcomings of the given connectionist<br />
model <strong>and</strong> sketch a direction for future research.<br />
2. Situated embodied semantics, grounding <strong>and</strong> neural nets<br />
For the sake of reference, rather than with the ambition of introducing one more<br />
“semantic theory” into the field, I call the approach to explaining linguistic<br />
meaning that I am pursuing situated embodied semantics (cf. Zlatev 1997). In<br />
brief, it emphasizes the following aspects of human language.<br />
Language is situated, i.e. interwoven into the practices in which it is used;<br />
it takes place in the world, it is not just about the world, i.e. not a “picture of<br />
reality”. Situatedness requires interaction with an environment through sense<br />
<strong>and</strong> motor organs, i.e. embodiment. Learning a language is achieved through<br />
active participation in practices, “language games”, <strong>and</strong> much of this participation<br />
can be seen as “training”. Knowing a language is practical mastery<br />
(know-how), not the internalization of an underlying system (“la langue”,<br />
“competence”). To know the meaning of an expression is thus to be able to use<br />
it in appropriate situations.