martes, 1 de febrero de 2011

Ray Jackendoff

Tuve noticia por primera vez de Ray Jackendoff en 1991, al finalizar el Cours de Spécialisation en la UNI II de Ginebra, al que asistí como oyente. La codirectora del curso, Maghie King, me recomendó el último libro que por entonces había publicado ese autor, Semantics and Cognition, que encontré muy interesante. Contenía análisis muy intuitivos de ciertas áreas de conceptos, aunque se echaba en falta una formalización única que permitiera concebir un modelo operable. Aun así, para mí fue un alivio ver por primera vez análisis semánticos ilustrados con gráficos, en lugar de la típica descomposición generativista de to kill, sacada directamente de la chistera del autor de turno. (Por cierto, ¿por qué utilizan tan a menudo los lingüistas el verbo kill en sus ejemplos?)

Hace pocas semanas, veinte años después, estando ya bastante avanzado el texto de mi libro, decidí pedirle ayuda. Después de haberlo intentado (infructuosamente) con Chomsky, Langacker y Talmy, esta vez era el turno de Jackendoff. Estaba pues redactando cuidadosamente el email que pensaba enviarle cuando, a falta sólo de la última palabra, toqué una tecla equivocada y el mensaje salió como una flecha hacia la Tufts University. El mensaje decía así:

"Dear Mr. Jackendoff,

I am not a linguist, but a theoretical physicist. For a long time, though, I have been working on certain aspects of language from the standpoint of information. A qualitative theory of information seems to be lacking in the literature, as well as a systematic, objective formalisation of the topological aspects of semantics. Your attempts in that direction are, in my view, the most rigorous undertaken to date.

That is why I think your criterion would be invaluable to me. Unfortunately, I am not in contact with any linguists and, therefore, I have hardly had the opportunity to test how sound or flawed my ideas may be. Besides, I do not think that just any linguist would feel comfortable with an unorthodox approach of language that bases syntax on information and semantics on topology.

I can imagine how much you value your time, so I would not dare ask you to do a thorough review of my work. I would merely ask you to read, from a critical perspective, a few pages of the paper I am writing. After working in isolation for 25 years, I really need some help, even if only to know whether I am treading solid ground or just .

Thank you very much in advance,
C. S."

Tal vez no había sido un error, sino que era mi inconsciente el que me había empujado a apretar la tecla de envío. Al darme cuenta de mi error, redacté un segundo email, que le envié al día siguiente, con el texto siguiente:

"Sorry. I accidentally sent you an email yesterday while I was editing the last word of the message (which might have been 'out of my mind' or something of the sort). Actually, I wanted to attach for you an abstract (three paragraphs) of my work, which is what I am doing now. It would be a great help if you could spare a couple of minutes to read it, and just let me know if you think it has any interest at all.

Thank you for your patience.
C. S."

Yo sabía que el adjunto era demasiado breve, pero tal vez Ray Jackendoff no tenía tanto tiempo para ocuparse de ese tipo raro que se había colado en su bandeja de entrada. Mis posibilidades eran pocas y, para aprovecharlas al máximo, mi idea había sido presentarle un cebo. Si aquellos tres párrafos conseguían despertar su curiosidad, se interesaría por el texto completo. Fue una apuesta. El 'abstract' que le adjunté decía así:

"¿How do languages convey information? An information process can be defined as a series of choices, which in turn imply a number of categories to choose from. An investigation of categories and their combinations as a means for disambiguation leads to a number of objects and operations that can be seen as the basic elements of syntax and predication. Furthemore, categories can be seen to involve underlying structures that can also be formally defined in terms of their denotation potential. From the standpoint of adjacency, such structures exhibit quasi-topological properties, which are rather convenient for the compression of information as well as for accommodating new information items.

Actually, this is what should be expected. The ultimate input for language are sensory perceptions of a very different nature. Therefore, it should be possible to refer them to some common format before integrating them into a one-dimensional output such as language. The common format proposed here consists of a number of denotational configurations which, by their geometric nature, are irreducible, and would seem ideal as semantic primitives. With an absolute referent for meaning and a symbolic formalism for syntax, linguists’ hypotheses could eventually be tested, and linguistics could hopefully become a real science.

Natural language categories and their associated semantic structures seem to provide a good means to unify syntax and semantics, at the same time combining the structured categories in clusters of semantic representations. To illustrate the implications of the model, my work includes a discussion on the semantics of indefinites and a few hints at its broader implications in natural language and set theory.

[If you are interested, I can elaborate.]"

Casi a vuelta de correo, Jackendoff me respondió:

"Dear Mr. S.,

Thank you for your inquiry.  I find your abstract rather difficult to understand.  However, I tend to be skeptical of formal approaches of the sort you seem to have in mind.  My sense is that formalism in linguistic theory is ultimately at the service of psychology and theories of neural computation. The first chapter of my book Foundations of Language lays out some formal criteria that I think any theory of language must satisfy, and the last section of chapter 3 puts this in the context of neural computation.

I'm sorry I can't be more helpful.  I wish you all the best in your research.

Ray Jackendoff"

Ese "I find your abstract rather difficult to understand", más que ninguna otra frase, me desmoralizó. Releí mi resumen. Efectivamente, estaba demasiado condensado. Tanto, que la apuesta no había dado resultado. Al principio, no supe cómo reaccionar. En caliente nunca he sido buen argumentador. Escribí y borré varias respuestas diferentes, transcurrió un rato, releí su mensaje por enésima vez. Y entonces se hizo la luz. ¡Casualmente, el libro al que Jackendoff me remitía estaba en mi biblioteca! Lo compré el verano pasado. Había leído incluso unos cuantos capítulos antes de abandonarlo, decepcionado. Jackendoff ha abandonado aquella interesante línea de investigación de los años 80 para caer en brazos de la ptolomeica teoría chomskyana de Government-Binding.

Busqué el libro, lo abrí y me fui directamente al Capítulo 1. Yo también encuentro estos textos de mi interlocutor "rather difficult to understand", pero la última sección del capítulo (Anaphora and unbounded dependencies), acompañada de abundantes ejemplos, se entiende perfectamente. Lo que enturbia la visión de Jackendoff, de Chomsky y de la inmensa mayoría de los lingüistas es lo que yo aprendí traduciendo textos de telefonía digital en la UIT, en los años 80. En lugar de explicarlo aquí, reproduzco el mensaje con que contesté a Jackendoff, también casi a vuelta de correo:

"Dear Mr. Jackendoff,

Thank you very much for your reply. In the first chapter of your book Foundations of Language, the term "conditions" refers rather to *conventions* implicitly established by the users of languages. However, if the sentence (b) below is the agreed way to inquire about the object of the verb in (a):

(a) Beth ate bread
(b) What did Beth eat?

there is no reason for the second sentence below to be ungrammatical:

(1) Beth ate peanut butter and bread
(2) What did Beth eat peanut butter and for dinner?

except for the fact that it has never been used (before you), and is therefore a construction not expected by the receiver. Besides, (2) is the *only* way to ask about the word 'bread' in (1). A remedial construction such as:

Beth ate peanut butter and what?

follows the rule implied by

Beth ate what?

and not the rule used to consistently construct (b) and (2). An important thing to be aware of is the fact that languages are incomplete and conventional. Languages evolve, and not only morphologically. In Footnote 2 to Chapter 5 of his "Syntactic Structures", Chomsky wrote in 1975 "...many would question the grammaticalness of, e.g., 'John enjoyed and my friend liked the play'". Such constructions are nowadays generally accepted, as were passive English forms at some point in time, but not before the end of the eighteenth century. More complex constructions such as "he was given a book" were also for a long time inexistent and, therefore, deemed ungrammatical in the past.

The incompleteness of natural languages is an essential fact that may make appear as objective certain concepts about language that are actually subjective or merely conventional. Human languages are very effective compressors of information, which makes formal syntax relatively irrelevant, except as a tool to disambiguate and predicate by combining individual words and word groups. Expressions such as "place cheap eat" (as asked, for example, by a foreigner) are perfectly understandable English. They are still language, and they can be analysed in terms of information, but not so much as a grammatical production, whatever the grammar rules may be, given that such rules can be almost arbitrarily changed.

That different approach of you may be the reason for your skepticism. To further elaborate on my arguments, I am attaching now a few initial pages of my work, in the hope that you might be interested. I intend to send the whole paper to some journal for publication, and I would be immensely grateful for any critical reading of it.

Again, thank you for your patience.
C. S."

Veinticuatro horas después de enviarlo, aún no he recibido respuesta. ¿La recibiré? Se admiten apuestas.

No hay comentarios:

Turbo Tagger