\documentclass[11pt,a4paper]{article}
\usepackage[pdftex]{graphicx}
\usepackage[english]{babel}
\usepackage[pdfborder={0 0 0}]{hyperref}
\usepackage{verbatim}
%\usepackage[cm]{fullpage}
\usepackage[left=2cm,top=2cm,right=2cm,bottom=2cm]{geometry}
\usepackage{setspace} \onehalfspacing
\usepackage{sober}
%\usepackage{times}
%\pagestyle{plain}

\begin{document}

\begin{center}
{\Large {\bf Plausibility considered as a research strategy}: \\
	The Poverty of Plausibility }

{Andreas van Cranenburgh\footnote{\texttt{andreas@unstable.nl}}, \today} \\
{\em Essay proposal for Language \& Cognition course, University of Amsterdam}
\end{center}

%\abstract{ }
\tableofcontents

\section{Research question, context}
Is the desideratum of plausibility in Cognitive Linguistics actually a bias
which merits apprehension? If so, what could form a sound alternative to this
strategy? 

Cognitive Linguistics, as conceived by Langacker (1987) should strive for
``naturalness in linguistic description,'' or plausiblity according to Lakoff
and Johnson (1999). This constitutes a certain bias to account for the facts at
hand, as opposed to, say, the desideratum that a theory provide a working
demonstration or explanation of an ability, such as with computational models
of language (Klein, 2005). \footnote{Interestingly, Klein (2005) explicitly
mentions plausibility in its abstract:
\begin{quote}
[...] while the models in this work are not intended to be cognitively
plausible, their effectiveness can inform the investigation of what biases are
or are not needed in the human acquisition of language.
\end{quote}
}

Such approaches usually substitute wonderous abilities (eg., language
understanding) with more general, yet equally unaccounted for primitives,
such as intention-reading (Tomasello, 2003) and perceptual similarity for
categorization (Lakof \& Johnson, 1999). Humans ostensibly do, of course, read
intentions, just as they understand language; but unless this ability can be
modelled, introducing new primitives is mere hand waving. This is reminiscent
of associationism (Locke, Hume, etc.) which is decomposable in three faculties:
similarity, contiguity and detecting cause and effect. This is close to the
way generalization is declared to work in cognitive grammar. Associationism
seemed plausible enough, but turned out to have no bearing on actual
physiological processes aside from accurately describing (but not quite
explaining) conditioning.

Cognitive linguistics also relies on Gestalt psychology, which suffers from
similar problems:

\begin{quote}
	``The physiological theory of the Gestaltists has fallen by the
	wayside, leaving us with a set of descriptive principles, but without a
	model of perceptual processing. Indeed, some of their `laws' of
	perceptual organisation today sound vague and inadequate. What is meant
	by a `good' or `simple' shape, for example?'' (Bruce et al., 1996) 
	as quoted in Wikipedia article on Gestalt psychology.
\end{quote}

%Gestalt Psychology (used by Langacker) postulate unexplained faculties as
%primitives, such as `the gestalt effect' etc.
% - Principle of psychophysical isomorphism
%	unrealiability of introspection
% - Biotic experiment
%	impossible to separate influences of context with a control subject
%	not in that context (because that would not be a biotic experiment).
% "descriptive rather than explanatory in nature."

%minimal experimental interference
%	- but the point of science is to predict something will happen and then
%	  demonstrate it: interference.


\section{Elaboration, projected contents}

\subsection{Plausibility as bias}
\addcontentsline{toc}{subsubsection}{\numberline {2.1.1}Speculating to plausibility}
\addcontentsline{toc}{subsubsection}{\numberline {2.1.2}Eliminative materialism versus folk psychology}
\addcontentsline{toc}{subsubsection}{\numberline {2.1.3}Explaining the irrelevant: epiphenomena}

In an obvious sense plausibility is a terrible bias. Would humanity have
discovered lasers, X-rays and atom bombs if we held on to such a principle?
Not to mention imaginary numbers, infinite dimensional (Hilbert) spaces and
space-time curvature. Concepts accepted by science can be thoroughly
implausible, yet especially such concepts have revolutionized our understanding
of the world.

Language is of course another matter, being invented though not yet understood.
Still, an insistence on dealing with the facts at hand instead of constructing
an idealization is unwarranted. The facts present a very limited view of the
actual workings of the phenomenon, being derived from introspection or
hopelessly limited interpretations of neural activity. Dennett (2007) sums it
up nicely:

\begin{quote}
``Anybody who thinks that philosophers have found a method of {\em grammatical}
inquiry that is somehow immune to [...] the problems that can arise for that
anthropological inquiry owes us an apologia explaining just how the trick is
turned.'' (emphasis in the original)
\end{quote}

What ever happened to the venerable admonition that ``appearances can be
deceived''? The focus on overt phenomena is limiting.

Cognitive linguistics glorifies our common sense understanding of language. It
seems to wholly depend on folk psychology being correct about the mental 
world. This is certainly not something to be taken for granted, as any
physicalist understanding of the world should, arguably, contradict the
existence of causal mental states (Feyerabend, 1963). Just as our folk physics
and folk arithmetic proved to be radically false (or wanting, in any case), so
could (should?) folk psychology.


With any complex process epiphenomena are bound to occur. Hofstadter (1979)
gives the example of a computer system that crashes as soon as seven
programs are running concurrently. This does not mean that to fix this problem
one can simply go and look for instances of the number 7, be it in the code or
in the structure of the hardware, the number 7 is simply an accidental,
emergent property. Analogously, things like processing time for sentence
comprehension are probably epiphenomena, which a linguistic theory need not
account for. There is a methodology which happily abstracts over such
side-effects, it is one of formal modelling.


%\subsection{Desideratum or dogma?}

\subsection{Psychologism and the dismissal of formal methods}
\addcontentsline{toc}{subsubsection}{\numberline {2.2.1}The new psychologism of language}
\addcontentsline{toc}{subsubsection}{\numberline {2.2.2}Arguments for the irrelevance of formal methods}
\addcontentsline{toc}{subsubsection}{\numberline {2.2.3}Synthesis}

Cognitive linguistics brings with it a resurgence of psychologism, and
with that an apprehension of anything formal. Langacker (1987, p.\ 440) has
this to say about formal methods (relegated to a footnote!):

\begin{quote}
``If a language were regarded as a formal system, we could say that [structure]
(S) is algorithmically derivable. I am not so much opposed to this way of
viewing things as I am concerned with giving it a plausible cognitive
interpretation. Still, I would argue that a completely formal, algorithmic
account of linguistic expression is achievable only at the expense of imposing
arbitrary boundaries.''
\end{quote}

Much of the plausibility-motivated theories rail against formal methods. This
is a very unfortunate move, because formal methods can very well be adapted to
new findings, and introduce much needed rigor and systematicity. Van Benthem
(2008) argues cogently for maintaining both logic and psychologism:

\begin{quote}
``Advertising `mismatches' between inferential predictions of logical systems,
usually without proper attention to the modeling phase, and what is observed
in experiments with human subjects seems entirely the wrong focus to me -- not
to mention the fact that it is silly and boring.''
\end{quote}

Logic is about defining what is true, so it can be anything that is necessary
to arrive at true propositions. Saying there is no logical structure to
language is like saying language is not about anything in particular!


\subsection{Philosophy as an afterthought?}
\addcontentsline{toc}{subsubsection}{\numberline {2.3.1}What is grammar?}
\addcontentsline{toc}{subsubsection}{\numberline {2.3.2}Leaving theory at the laboratory door?}

For Langacker (1987), grammar consists of a pairing of utterances and their
meanings (ie., a set of associations). Lakoff and Johnson (1999, p.\ 498) are
more precise:

\begin{quote}
``The grammar of a language consists of the highly structured neural
connections linking the conceptual and expressive (phonological) aspects of the
brain'' 
\end{quote}

Both of these definitions miss the point. These definitions amount to saying
that the moon is made of bits because a computer image of the moon happens to
consist of pixels. A particular representation of language, be it in the brain
or in language use, does not constitute language in any useful (explanatory)
sense. An implementation is not an algorithm, just as a cake is not a recipe.
Perhaps psychologists can't have their proverbial cake and explain it too! 

% No, an image of the moon is not constituted by a set of pixels, since the
% moon demonstrably does not consist of pixels.

%fragment from email, 17-7-2009, (neurocogling-bounces@polymathix.com):
%What I was trying to point out in my quote is that a particular
%implementation of language, ie., someone's language performance, is not
%language proper, it is the phenomenon of some immaterial and elusive
%noumenon, language-in-itself (the same point applies to grammar, they can't
%be used interchangingly but it doesn't matter for the point I'm making) 
%
%Do you know about Hans the horse, who seemed to be able to do arithmetic?
%It turned out he used subtle cues of his caretakers to determine the right
%answer (these cues were unintentional).
%
%If you wouldn't know about this fact you might believe Hans was able to do
%arithmetic. However, no amount of arithmetic exercises is going to prove Hans
%can or can not do arithmetic (eg., when he fails the excersize might
%plausibly be too difficult). Neither is it going to help to put him in a
%brain scanner, perhaps horses have quite different ways of legitimately doing
%their arithmetic?
%
%The question of how one *can* prove someone does arithmetic (or uses
%language) I leave as an excersize to the reader, but it should be clear that
%behaviorism is out (or at least severely deceptive, Hans was quite the rage
%in his age) ...


Without a guiding principle as to what to look for, it is impossible to find
something. Lakoff and Johnson (1999) insist on the necessity of doing away with
what they call ``a priori philosophy,'' but what about plausibility? That seems
a perfectly {\em a priori notion}, a bias that theories should conform to one's
preconceived worldview.

It seems that cognitive linguistics will be applicable to any kind of data,
something which suggests unfalsifiability (just as the UG, incidentally).
This means the Texas sharpshooter fallacy\footnote{``the fallacy of selecting
or adjusting a hypothesis after the data is collected, making it impossible to
test the hypothesis fairly. Refers to the concept of firing shots at a barn
door, drawing a circle around the best group and declaring that to be the
target.'' (Wikipedia, List of cognitive biases)} is being invoked. It
should be a universal requirement of linguistic theories to make testable
predictions.  For example, a theory might propose the most efficient strategy
for learning a language. What is needed is a theory that does for language what
David Marr did for visual perception:

\begin{quote}
``For Marr, no psychological study that is not based on prior abstract
task-analysis can be more than mere anecdotal reportage or idle speculation."
 (Boden, 1990)
\end{quote}

Perhaps it is time for a Computational Positivism.\footnote{Much to my dismay,
a google search pointed out that this idea has already been coined (Narasimha,
2003); so much for creativity in the internet age. On the other hand, it is
vindicating that the approach rests on solid historical foundations, viz. the
desire to produce algorithms concordant with observation.} This is not meant
to suggest a return to the Chomskian paradigm; developments like the Neural
Theory of Language (Feldman, 2008) seem to provide a more fertile ground,
although it is unclear how formal such an approach can be.

%17-7-2009 comment: clarify that I'm not in favor of computationalism (the
%mind is not a computer), but a computational model is still very useful in
%bolstering claims about cognition.
%Also: NTL is not neural. Perhaps re-word: unclear if it works at all when
%formalized!

%Conscious versus unconscious learning:
%During critical period learning is conscious, but further intellectual
%development is conscious (thinking critically).
	%"Learning how to think" really means learning how to exercise some
	%control over how and what you think. It means being conscious and
	%aware enough to choose what you pay attention to and to choose how you
	%construct meaning from experience. (David Foster Wallace,
%	http://online.wsj.com/article/SB122178211966454607.html)
\section{Expected conclusion}
The study of language is a difficult endeavour fraught with luring temptations
and simplifications. A faithful method of keeping linguistics in check should
be the requirement of providing demonstrable, computational theories, not a
bias for plausibility (which is ill-defined and arbitrary). This has the added
benefit that formal methods can be re-united with psychologistic
considerations, bridging the gap between accounts of competence and empirical
data of performance. The ultimate scope of formal and neural methods in
explaining language remains to be demarcated.

\section{Bibliography}
%turn off bold for labels (all bow to magic incantations!)
%\renewcommand{\descriptionlabel}[1] {\hspace{\labelsep}#1}

\begin{description}

\item[Benthem, Johan van] (2008), ``Logic and Reasoning: do the facts matter?''
Studia Logica vol.\ 88: pp.\ 67-84. 

\item[Boden, Margaret A.] (1990), ``Has AI helped psychology?,'' in ``The
foundations of artificial intelligence,'' eds.\ Derek Partridge \& Yorick
Wilks, Cambridge University Press.

\item[Bruce, V., Green, P. \& Georgeson, M.] (1996), ``Visual perception: Physiology, psychology and ecology,'' 3rd ed., Psychology Press.

\item[Dennett, Daniel] (2007), ``Philosophy as naive anthropology'' in
``Neuroscience and Philosophy: Brain, Mind, and Language,'' co-authored with
Maxwell Bennett, Peter Hacker, and John Searle. Columbia University Press.

\item[Feldman, Jerome A.] (2008), ``From Molecule to Metaphor: A neural theory
of language,'' MIT press paperback edition.

\item[Feyerabend, Paul] (1963), ``Materialism and the Mind-Body Problem,'' Review of Metaphysics vol.\ 17: pp.\ 49-66.

\item[Hofstadter, Douglas] (1979), ``G\"odel, Escher, Bach: An Eternal Golden Braid,'' Basic Books.

\item[Klein, Dan] (2005), ``The Unsupervised Learning of Natural Language Structure,'' Dan Klein, Ph.D. Thesis, Stanford University. 
\url{http://www.cs.berkeley.edu/~klein/papers/klein_thesis.pdf}


\item[Lakoff, George \& Johnson, Mark] (1999), ``Philosophy In The Flesh: The
Embodied Mind and Its Challenge to Western Thought,'' Basic Books.

\item[Langacker, Ronald] (1987). ``Foundations of Cognitive Grammar, Volume I, Theoretical Prerequisites.'' Stanford, California: Stanford University Press.
 
%(1998), ``Conceptualization, symbolization and grammar,'' In Tomasello, Michael. 1998, ``The New Psychology of Language: Cognitive and functional approaches to language structure,'' Mahwah, NJ: Laurence Erlbaum.


%\item[Montague, Richard] (1973), ``The Proper Treatment of Quantification in
%Ordinary English,'' reprinted in ``Formal Semantics: The Essential Readings,''
%by Paul Portner, Barbara H. Partee, eds. %Blackwell, 2002. ISBN 0631215425

\item[Narasimha, Roddam] (2003), ``Axiomatism and Computational Positivism,''
Reprinted from Economic and Political Weekly, vol.\ XXXVII no 35, pp. 3650-3656 \\
\url{http://www.indianscience.org/reviews/13-%20R--Roddam-%20Axiomatism%20-%20checked%20by%20pankaj.pdf}

%\item[Pereira, Fernando] (2000), ``Formal grammar and information theory:
%together again?'', Philosophical Transactions of the Royal Society 358(1769):
%1239-1253.
%\url{http://www.cis.upenn.edu/%7epereira/papers/rsoc.pdf}

%\item[Putnam, Hilary] (1975), ``The meaning of `meaning','' in Language, Mind and Knowledge, ed.\ K. Gunderson

\item[Tomasello, Micheal] (2003), ``Constructing a Language. A Usage-Based Theory of Language Acquisition,'' Cambridge MA: Harvard University Press.

\end{description}
\end{document}


%http://www.springerlink.com/content/k26j8675jl448208/

%[on construction grammar]
%The earliest study was "There-Constructions," which appeared as Case Study 3
%in George Lakoff's Women, Fire, and Dangerous Things.[1] It argued that the
%meaning of the whole was not a function of the meanings of the parts, that odd
%grammatical properties of Deictic There-constructions followed from the
%pragmatic meaning of the construction, and that variations on the central
%construction could be seen as simple extensions using form-meaning pairs of
%the central construction.

%Searle: syntax by itself is not constitutive of semantics.
% --- This point is missed so often, it bears repeating: the syntactically specifiable objects over which computations are defined can and standardly do possess a semantics; it's just that the semantics is not involved in the specification.
%Rey (2002) attacking Searle
%this means that characterizing programs as formal systems is unfair
%a program needs semantic specification as well.
%Rey, G., 2002, Searle's Misunderstandings of Functionalism and Strong AI  in Preston and Bishop (eds.)
%Searle, J., 1980, ‘Minds, Brains and Programs’, Behavioral and Brain Sciences, 3:417-57

%[..]there is, of course, a familiar philosophical ambiguity lurking in the
%wings here -- the confusion of behaving in a fashion describable by a rule
%with following or applying a rule -- and arguably advocates of an algorithmic
%level of description have not always kept this distinction in mind. 
%On Marr's computational theory of vision

%However, Lowenheim-Skolem assures that there is at least one interpretation S*
%that maps all of the referring terms onto only mathematical objects. S* cannot
%be the canonical interpretation, but there is nothing in the syntax of
%Mentalese to explain why S is the correct interpretation and S* is not.
%Therefore syntax underdetermines semantics. (Compare acknowledgement of this
%in Pylyshyn 1984: page 44.)
%(criticism that representational theory of mind does not account for semantics)