I think Bühler's work is interestin; and while I don't necessary agree with it philosophically, it is nice that he never left the human behind in a web of theory. It was quite the opposite.
I like this guy more than Frege or Montague. Still, please avoid overburdening language with what it does not do - a key doesn't perform the functions of value. Your pushing the acceleration pedal doesn't move the car, it sends a signal, starts a chain of events that result in the car moving.
Thanks for your comments. I have read your article on pointing and I am afraid that I was disappointed by it. There are lots of elements I would expect see that are not present. I was thinking of drafting a point by point reply but I fear lack the time.
If I may point to the section on Ambiguity wich does not address research in psycho/neurolinguistics and the dismabiguation mechanisms and real time resolution of ambiguities.
Or I would very much like to understand how your concept of "pointing" accounts for language acquisition. Extensive research by Tomasello (2003), Bruner (1983), and others demonstrates that language acquisition occurs through social interaction and shared attention mechanisms. Children learn language not merely by having words "point" to objects but through complex social-cognitive processes.
Children's early language includes not only referential terms (which might support a "pointing" function) but also social regulatory expressions, emotional expressions, and ritual language. The article's narrow focus on pointing does account for the diversity of functions in early language development.
I think a lot of the problem is that your idea on "pointing" is introduced without defining the term in any formal or testable way. While ‘pointing’ is an intuitive metaphor, the article does not distinguish between deixis (a well-established linguistic concept involving spatial, temporal, and personal reference) and broader claims about reference, indexicality, or symbolic behaviour. As a result, the core thesis lacks precision and risks collapsing into triviality.
My apprach may be far too rigid and "academic", but there is value in a methodology tht demands that terms are defines, that statemetns are backed up and that a frame of reference is agreed, and, if a new challenging frame, it is presented, explained and defined in a way with validates non using past and present knowledge.
There are many phenomena that can be pointed - property (like color), object (like apple), action (like give), abstraction (like happiness), state (like tired), facts (like "I give you a red apple"), rules (like "touching fire results in burns"), etc. All those can be figured out based on context but require different amount of data/repetitions. Respectively the more difficult ones take longer to acquire - just as observed.
Deixis and many other linguistic phenomena are also easy to explain - they stand for references to previously referenced types - objects, people, time periods, locations. Rigidity/precision hardly apply to language and the meaning of words, they are always subject to "figuring out" (like in cooperative completion).
If you are interested, I can develop a more thorough text. Or even better I could do that in collaboration - do you know anyone who may be interested in such a joint project?
I find it difficult to get where you are trying to go. I caanot find a solid ground, just floatinf words and concepts. Maybe I am too stuck within an academic framework, but having clear definitions, clear theories which can then be challenged is, for me, the best option.
I think your reply unfortunately reinforces some of the key problems in the original article. It conflates different linguistic phenomena without clarifying their distinctions, and it continues to rely on metaphor and intuition rather than conceptual precision or theoretical engagement. Let me address the points in turn:
1. "Many phenomena can be pointed at..."
You suggest that properties, objects, actions, abstractions, states, facts, and rules can all be “pointed at.” That is a generous interpretation of the term pointing, but it raises two serious problems:
- Conceptual overstretch: If everything can be said to be "pointed at," then the term ceases to differentiate between types of linguistic reference. The semantic and cognitive mechanisms behind referring to “a red apple,” “happiness,” or “rules” are vastly different. Reference to concrete objects often relies on perceptual grounding or deixis, while abstract reference involves conceptual generalisation and may not be indexical at all. Using pointing to cover both collapses distinctions that are foundational in semantics, pragmatics, and cognitive linguistics.
- Theoretical emptiness: Pointing becomes a metaphor that explains everything and therefore nothing. It is not enough to say that "rules can be pointed at." How? Under what conditions? By which linguistic mechanisms?
2. “All those can be figured out based on context…”
Here, you seem to be alluding to the importance of context in interpretation, which is uncontroversial. But the claim that these meanings are simply “figured out” by repetition or exposure is too vague to be analytically useful.
- Are you referring to language acquisition? If so, what model are you drawing from—usage-based, statistical learning, generative theory, connectionism?
- Are you describing reference resolution in discourse? Then where is the engagement with theories like Relevance Theory, Discourse Representation Theory, or Centering Theory?
You make a general observation (“the more difficult ones take longer to acquire”) but offer no evidence, no data, and no acknowledgement that such issues have been rigorously investigated in developmental linguistics and psycholinguistics.
3. Deixis as “references to previously referenced types”
I see this as a mischaracterisation of deixis. Deictic expressions (e.g. this, that, now, here, you) do not always refer to “previously referenced types.” They often initiate reference, anchored in the speaker’s spatiotemporal context. Deixis is not merely anaphora or cohesion. It is context-sensitive reference that requires knowledge of the speech situation.
The crucial distinction between exophoric and endophoric deixis is also missing. I think h any serious discussion of deixis must acknowledge this distinction. Is it a lack of engagement with the literature?
4. “Rigidity/precision hardly apply to language…”
This is a claim that is too vague, borderline misleading. On one hand, yes, natural language is flexible and context-sensitive. On the other, many branches of linguistics are precisely concerned with how language achieves systematicity, predictability, and constraints. Consider:
- The rigidity of reference in Kripke’s modal logic (e.g. Hesperus and Phosphorus refer rigidly to Venus)
- The precision of compositional semantics in predicate logic
- The highly constrained syntax of natural languages, which supports predictability and grammaticality
Saying that language lacks precision overlooks the fact that many linguistic phenomena are tightly rule-governed and that this rule-governedness is what makes interpretation possible in the first place.
5. “Words are always subject to figuring out”
This appears to refer loosely to Grice’s notion of cooperative inference, but the phrasing is imprecise. Not all meanings are worked out afresh in every context. Much of language use relies on conventional meaning, stored in the lexicon, and governed by syntactic and morphological rules. Cooperative inference explains implicature, not the basic semantics of give, apple, or touch.
In other words, the fact that some pragmatic interpretation occurs does not mean that language has no stable structure or meaning.
I think, interestingly, that your writing asserts a general vision of language without acknowledging the theoretical, empirical, or methodological distinctions that make linguistic analysis possible. You use terminology from multiple subfields—acquisition, semantics, pragmatics—but without engaging their assumptions or findings. A philosophical or poetic perspective on language can be valuable, but, for me, it must be anchored in analytical clarity if it seeks to contribute to a (scholarly) conversation.
I fear that without a more careful and informed engagement with linguistic theory. your framework remains metaphorical, impressionistic, and circular.
To start with, I have my own explanatory theory of intelligence. It relies on the use of comparable properties and comparisons. The core algorithm of intelligence is the selection of the most fitting option from the available ones respecting relevant constraints. "Options" and "constraints" depend on the task. (In object recognition, options are categories, constraints are the object's properties. In a detective story, options are suspects, constraints are clues and alibis.)
My theory of language is based on the abovementioned theory. "Objects have properties, actions change properties." - it was my initial insight. Adjectives stand for primitive properties, nouns stand for sets of properties - fitting objects always have more properties.
Property in my theory is an axis with unique semantics. It is divided into ranges. These ranges enable interchangeability of fitting objects and differentiation of objects that fall in different ranges. Ranges are also known as Concepts. Boundaries between ranges are defining features of concepts.
Now, back to your comments.
"Conceptual overstretch: ... The semantic and cognitive mechanisms behind referring to “a red apple,” “happiness,” or “rules” are vastly different." Yes, different, and I mentioned that, but not clearly enough. Individual phenomena (abstract like happiness or specific like apple) may be referred to by noun phrases. Rules or facts combine objects and interactions. Referring to them requires sentences. A story requires multiple sentences to refer to it.
"You make a general observation (“the more difficult ones take longer to acquire”) but offer no evidence, no data, and no acknowledgement that such issues have been rigorously investigated in developmental linguistics and psycholinguistics."
"Deictic expressions (e.g. this, that, now, here, you) do not always refer to “previously referenced types.”
If anyone starts a conversation with any of those words alone and expects understanding, it will meet a clarifying question.
"The crucial distinction between exophoric and endophoric deixis is also missing."
It's an artificial distinction. All references are exophoric. Context in my theory is about phenomena, not words referring to them. Any text introduces context and then refers to it. Cataphoric reference introduces some pieces of information allowing to connect it to other pieces provided later.
"how language achieves systematicity, predictability, and constraints."
There are relevant pieces in a context the speaker wants to point to (be that story, fact, or phenomenon). There are ways (text with some structure, sentence with some grammar, noun phrase with some sequence of words) to achieve that. At the lowest level of references, what others call compositionality I call stacking filters based again on the use of comparable properties, comparisons and the core algorithm of intelligence.
"rigidity of reference in Kripke’s modal logic (e.g. Hesperus and Phosphorus refer rigidly to Venus)"
Do you know how many "Messi" boys play street football around the world? Do you know how many people have name Aristotle? Kripke did not know. Names follow the same filtering idea. That is why in a big crowd using the first name is not a good idea, neither is using the last name in a family.
"precision of compositional semantics" - it is achieved by stacking filters, not by composition of "meanings", which those authors cannot clearly explain. Meaning is what is pointed at, simple.
"highly constrained syntax of natural languages" - it is the continuation of the "convention idea" of de Saussure. We have conventions about meanings of words (I introduce there "referential flexibility" - it is easy to add new meanings but hard to expand that convention on larger communities, hence the smaller the community the more flexible they are about coming up with new slang or terms). We also have conventions about grammar - how facts should be referred to. We also have conventions about the structures of stories. At each level we have a range of options to select from based on specific needs (the core algorithm again).
"many linguistic phenomena are tightly rule-governed" - any rule has exceptions
"Grice’s notion of cooperative inference" - language may be cooperative or uncooperative. Interrogated criminals figure out how to misrepresent their private context in such a way that their story is believable and not criminal.
Yes, there are many conventions, like those about implicatures. But they have to be shared and the speaker needs to know for sure that listeners are aware of those. If I mention Harry Potter, a fan may recall the death of Dumbledore, but a person who did not read the books nor watched the movies will not produce that association.
LLMs literature misinterprets context to be "words", while I see context as phenomena and their interactions with history. It may be from memory or about one's beliefs, about abstractions (like in math) or imaginary (like Harry Potter) or hypothetical (like multiple worlds). I agree that poets do important work with respect to language, maybe I will try to address that in the future but at the moment I feel that my ideas need to be stated as clearly as possible. Any help is welcome. Your critical eye is so much appreciated. Thank you! Exactly what I and my theory need.
Alexander, thank you for engaging so extensively with the critique. You are generous in tone and clear about your intent, which is to develop an original theory of intelligence and language. That said, your reply confirms rather than resolves the problems raised earlier. What you are presenting is not yet a theory in any disciplinary sense, but rather a metaphorical framework that borrows freely from established domains without fully understanding or integrating them.
Let me try and respond to your points in turn.
1. "I have my own explanatory theory of intelligence..."
You begin with a statement that immediately requires scrutiny. The claim to possess a theory of intelligence carries significant weight. The term "theory" is not a placeholder for "framework I find intuitive." A theory, particularly in cognitive science or linguistics, must meet several conditions:
* Coherence: internally consistent definitions and operations
* Predictive power: ability to generate testable hypotheses
* Falsifiability: conditions under which it could be shown to be incorrect
* Integration: relation to existing research and empirical data
At present, your proposed "selection algorithm" that compares properties and selects options based on constraints is reminiscent of basic decision-making models in AI and cognitive science. Yet you provide no formalisation, no references to related models (e.g. ACT-R, Bayesian approaches, constraint-based grammars), and no empirical validation.
Calling it your own theory of intelligence does not grant it explanatory authority. It must earn that by engaging the existing literature and by making testable claims.
2. Objects, Properties, and Your Account of Language
Your theory of language is based on a reductionist schema: objects have properties; adjectives represent properties; nouns are sets of properties; actions change properties. While superficially attractive, this view is both under-theorised and overgeneralised.
* Properties as axes: You claim a property is an axis with unique semantics divided into ranges. This sounds like a rephrasing of feature-based semantics or dimensional models of meaning, such as those found in prototype theory or distributional semantics. Are you familiar with Rosch's categorisation work or Gärdenfors’ Conceptual Spaces? Without this context, your terminology remains disconnected from ongoing research.
* Concepts as ranges: The idea that concepts emerge from range boundaries is broadly compatible with theories in cognitive linguistics, but again, without citation, elaboration, or critical engagement, it floats as an assertion rather than an insight.
* "Actions change properties": This collapses a vast number of verb types, aspectual distinctions, and syntactic structures into a single relational schema. Vendler's classification of verb types, lexical aspect, and argument structure in syntax all offer much more nuanced and empirically supported accounts.
To claim explanatory power, a theory must do more than redescribe linguistic categories using new metaphors.
3. Your Dismissal of Linguistic Distinctions
Several responses in your reply point to a resistance to disciplinary concepts that appears more rhetorical than justified.
* "Exophoric vs endophoric deixis is artificial": This distinction is not a conceptual gimmick. It captures a crucial difference in reference resolution. Endophoric reference points within the text, while exophoric reference relies on situational context. Disregarding this collapses discourse analysis and reference pragmatics into a singular intuition about context.
* "Syntax is just convention": Reducing syntax to social convention ignores decades of empirical work showing that syntactic constraints are not only cross-linguistically patterned but also cognitively processed with remarkable consistency. It also ignores work on universals, processing models, and syntactic priming. Rules may have exceptions, but the presence of exception does not abolish the rule.
* "Compositionality is stacking filters": You reject the compositionality principle without understanding what it seeks to explain. Compositionality in semantics is not about meaning per se, but about systematicity—how complex meanings depend on the meanings of parts and their arrangement. Replacing this with a vague idea of filters again trades one metaphor for another without solving anything.
4. Empirical Claims without Evidence
You write: "Do I have to? I thought it is a well-established fact..." Yes, you do have to! If you are proposing a new theory, especially one that crosses disciplinary boundaries, you are responsible for grounding it in existing knowledge or showing where and why that knowledge is insufficient. Simply referring to a child development website is not adequate in a scholarly exchange.
If your theory involves language acquisition, then cite Tomasello, Clark, Gleitman, or Karmiloff-Smith. If it involves the relationship between perception and language, cite Jackendoff or Barsalou. Otherwise, it remains anecdotal.
5. On Names, Reference, and Rigidity
You attempt to counter Kripke’s theory of rigid designation by pointing out that many people share the name “Messi” or “Aristotle.” But this is a misunderstanding of Kripke’s point. Kripke’s theory is not about uniqueness, but about how a name refers, namely through a causal-historical chain. That some names are ambiguous in use is a problem of disambiguation, not of reference rigidity.
Your counterexample does not undermine the theory. It just misrepresents it.
6. Final Points: Context, Poetry, and Claims to Innovation
You end by distinguishing your own theory of context from that of LLMs and mention that poets do important work with language. These gestures are not unwelcome, but they remain vague. Your notion of context as "phenomena and their interactions with history" is interesting, but again, lacks formal content. Are you proposing an event-based ontology? A mental model framework? Are you aware of situation semantics or discourse representation theory?
More importantly, if you genuinely want to develop a new theory, you must accept the burdens that come with that ambition:
* You need to define your terms clearly and consistently.
* You need to show how your theory differs from existing ones, and why that matters.
* You need to show what your theory explains that others cannot.
* You need to be able to show where it can be tested, challenged, or extended.
Otherwise, what you have is not yet a theory. It is a set of personal intuitions expressed in idiosyncratic language, with metaphor standing in for argument.
Thank you for these comments. Now I know how I should present my ideas. I am about to write another post or two. I hope you will agree to take a look and comment. I find your attitude and expertise extremely helpful.
Thank you so very much
Interesting — I suppose that explains the failure of Yale professor Pierre Capretz’s method of teaching French by “pointing.”
https://www.openculture.com/2011/11/french_in_action_cult_classic_french_lessons_from_yale.html
I think Bühler's work is interestin; and while I don't necessary agree with it philosophically, it is nice that he never left the human behind in a web of theory. It was quite the opposite.
I like this guy more than Frege or Montague. Still, please avoid overburdening language with what it does not do - a key doesn't perform the functions of value. Your pushing the acceleration pedal doesn't move the car, it sends a signal, starts a chain of events that result in the car moving.
Also, the role of language is good but the mechanism is also important. This post describes not only what I view as the role of language but also how it achieves that https://alexandernaumenko.substack.com/p/the-pointing-role-of-language
Thanks for your comments. I have read your article on pointing and I am afraid that I was disappointed by it. There are lots of elements I would expect see that are not present. I was thinking of drafting a point by point reply but I fear lack the time.
If I may point to the section on Ambiguity wich does not address research in psycho/neurolinguistics and the dismabiguation mechanisms and real time resolution of ambiguities.
Or I would very much like to understand how your concept of "pointing" accounts for language acquisition. Extensive research by Tomasello (2003), Bruner (1983), and others demonstrates that language acquisition occurs through social interaction and shared attention mechanisms. Children learn language not merely by having words "point" to objects but through complex social-cognitive processes.
Children's early language includes not only referential terms (which might support a "pointing" function) but also social regulatory expressions, emotional expressions, and ritual language. The article's narrow focus on pointing does account for the diversity of functions in early language development.
I think a lot of the problem is that your idea on "pointing" is introduced without defining the term in any formal or testable way. While ‘pointing’ is an intuitive metaphor, the article does not distinguish between deixis (a well-established linguistic concept involving spatial, temporal, and personal reference) and broader claims about reference, indexicality, or symbolic behaviour. As a result, the core thesis lacks precision and risks collapsing into triviality.
My apprach may be far too rigid and "academic", but there is value in a methodology tht demands that terms are defines, that statemetns are backed up and that a frame of reference is agreed, and, if a new challenging frame, it is presented, explained and defined in a way with validates non using past and present knowledge.
There are many phenomena that can be pointed - property (like color), object (like apple), action (like give), abstraction (like happiness), state (like tired), facts (like "I give you a red apple"), rules (like "touching fire results in burns"), etc. All those can be figured out based on context but require different amount of data/repetitions. Respectively the more difficult ones take longer to acquire - just as observed.
Deixis and many other linguistic phenomena are also easy to explain - they stand for references to previously referenced types - objects, people, time periods, locations. Rigidity/precision hardly apply to language and the meaning of words, they are always subject to "figuring out" (like in cooperative completion).
If you are interested, I can develop a more thorough text. Or even better I could do that in collaboration - do you know anyone who may be interested in such a joint project?
I find it difficult to get where you are trying to go. I caanot find a solid ground, just floatinf words and concepts. Maybe I am too stuck within an academic framework, but having clear definitions, clear theories which can then be challenged is, for me, the best option.
I think your reply unfortunately reinforces some of the key problems in the original article. It conflates different linguistic phenomena without clarifying their distinctions, and it continues to rely on metaphor and intuition rather than conceptual precision or theoretical engagement. Let me address the points in turn:
1. "Many phenomena can be pointed at..."
You suggest that properties, objects, actions, abstractions, states, facts, and rules can all be “pointed at.” That is a generous interpretation of the term pointing, but it raises two serious problems:
- Conceptual overstretch: If everything can be said to be "pointed at," then the term ceases to differentiate between types of linguistic reference. The semantic and cognitive mechanisms behind referring to “a red apple,” “happiness,” or “rules” are vastly different. Reference to concrete objects often relies on perceptual grounding or deixis, while abstract reference involves conceptual generalisation and may not be indexical at all. Using pointing to cover both collapses distinctions that are foundational in semantics, pragmatics, and cognitive linguistics.
- Theoretical emptiness: Pointing becomes a metaphor that explains everything and therefore nothing. It is not enough to say that "rules can be pointed at." How? Under what conditions? By which linguistic mechanisms?
2. “All those can be figured out based on context…”
Here, you seem to be alluding to the importance of context in interpretation, which is uncontroversial. But the claim that these meanings are simply “figured out” by repetition or exposure is too vague to be analytically useful.
- Are you referring to language acquisition? If so, what model are you drawing from—usage-based, statistical learning, generative theory, connectionism?
- Are you describing reference resolution in discourse? Then where is the engagement with theories like Relevance Theory, Discourse Representation Theory, or Centering Theory?
You make a general observation (“the more difficult ones take longer to acquire”) but offer no evidence, no data, and no acknowledgement that such issues have been rigorously investigated in developmental linguistics and psycholinguistics.
3. Deixis as “references to previously referenced types”
I see this as a mischaracterisation of deixis. Deictic expressions (e.g. this, that, now, here, you) do not always refer to “previously referenced types.” They often initiate reference, anchored in the speaker’s spatiotemporal context. Deixis is not merely anaphora or cohesion. It is context-sensitive reference that requires knowledge of the speech situation.
The crucial distinction between exophoric and endophoric deixis is also missing. I think h any serious discussion of deixis must acknowledge this distinction. Is it a lack of engagement with the literature?
4. “Rigidity/precision hardly apply to language…”
This is a claim that is too vague, borderline misleading. On one hand, yes, natural language is flexible and context-sensitive. On the other, many branches of linguistics are precisely concerned with how language achieves systematicity, predictability, and constraints. Consider:
- The rigidity of reference in Kripke’s modal logic (e.g. Hesperus and Phosphorus refer rigidly to Venus)
- The precision of compositional semantics in predicate logic
- The highly constrained syntax of natural languages, which supports predictability and grammaticality
Saying that language lacks precision overlooks the fact that many linguistic phenomena are tightly rule-governed and that this rule-governedness is what makes interpretation possible in the first place.
5. “Words are always subject to figuring out”
This appears to refer loosely to Grice’s notion of cooperative inference, but the phrasing is imprecise. Not all meanings are worked out afresh in every context. Much of language use relies on conventional meaning, stored in the lexicon, and governed by syntactic and morphological rules. Cooperative inference explains implicature, not the basic semantics of give, apple, or touch.
In other words, the fact that some pragmatic interpretation occurs does not mean that language has no stable structure or meaning.
I think, interestingly, that your writing asserts a general vision of language without acknowledging the theoretical, empirical, or methodological distinctions that make linguistic analysis possible. You use terminology from multiple subfields—acquisition, semantics, pragmatics—but without engaging their assumptions or findings. A philosophical or poetic perspective on language can be valuable, but, for me, it must be anchored in analytical clarity if it seeks to contribute to a (scholarly) conversation.
I fear that without a more careful and informed engagement with linguistic theory. your framework remains metaphorical, impressionistic, and circular.
To start with, I have my own explanatory theory of intelligence. It relies on the use of comparable properties and comparisons. The core algorithm of intelligence is the selection of the most fitting option from the available ones respecting relevant constraints. "Options" and "constraints" depend on the task. (In object recognition, options are categories, constraints are the object's properties. In a detective story, options are suspects, constraints are clues and alibis.)
My theory of language is based on the abovementioned theory. "Objects have properties, actions change properties." - it was my initial insight. Adjectives stand for primitive properties, nouns stand for sets of properties - fitting objects always have more properties.
Property in my theory is an axis with unique semantics. It is divided into ranges. These ranges enable interchangeability of fitting objects and differentiation of objects that fall in different ranges. Ranges are also known as Concepts. Boundaries between ranges are defining features of concepts.
Now, back to your comments.
"Conceptual overstretch: ... The semantic and cognitive mechanisms behind referring to “a red apple,” “happiness,” or “rules” are vastly different." Yes, different, and I mentioned that, but not clearly enough. Individual phenomena (abstract like happiness or specific like apple) may be referred to by noun phrases. Rules or facts combine objects and interactions. Referring to them requires sentences. A story requires multiple sentences to refer to it.
"You make a general observation (“the more difficult ones take longer to acquire”) but offer no evidence, no data, and no acknowledgement that such issues have been rigorously investigated in developmental linguistics and psycholinguistics."
Do I have to? I thought it is a well-established fact. For example, https://davincicollaborative.com/the-7-stages-of-language-acquisition-in-children/
"Deictic expressions (e.g. this, that, now, here, you) do not always refer to “previously referenced types.”
If anyone starts a conversation with any of those words alone and expects understanding, it will meet a clarifying question.
"The crucial distinction between exophoric and endophoric deixis is also missing."
It's an artificial distinction. All references are exophoric. Context in my theory is about phenomena, not words referring to them. Any text introduces context and then refers to it. Cataphoric reference introduces some pieces of information allowing to connect it to other pieces provided later.
"how language achieves systematicity, predictability, and constraints."
There are relevant pieces in a context the speaker wants to point to (be that story, fact, or phenomenon). There are ways (text with some structure, sentence with some grammar, noun phrase with some sequence of words) to achieve that. At the lowest level of references, what others call compositionality I call stacking filters based again on the use of comparable properties, comparisons and the core algorithm of intelligence.
"rigidity of reference in Kripke’s modal logic (e.g. Hesperus and Phosphorus refer rigidly to Venus)"
Do you know how many "Messi" boys play street football around the world? Do you know how many people have name Aristotle? Kripke did not know. Names follow the same filtering idea. That is why in a big crowd using the first name is not a good idea, neither is using the last name in a family.
"precision of compositional semantics" - it is achieved by stacking filters, not by composition of "meanings", which those authors cannot clearly explain. Meaning is what is pointed at, simple.
"highly constrained syntax of natural languages" - it is the continuation of the "convention idea" of de Saussure. We have conventions about meanings of words (I introduce there "referential flexibility" - it is easy to add new meanings but hard to expand that convention on larger communities, hence the smaller the community the more flexible they are about coming up with new slang or terms). We also have conventions about grammar - how facts should be referred to. We also have conventions about the structures of stories. At each level we have a range of options to select from based on specific needs (the core algorithm again).
"many linguistic phenomena are tightly rule-governed" - any rule has exceptions
"Grice’s notion of cooperative inference" - language may be cooperative or uncooperative. Interrogated criminals figure out how to misrepresent their private context in such a way that their story is believable and not criminal.
Yes, there are many conventions, like those about implicatures. But they have to be shared and the speaker needs to know for sure that listeners are aware of those. If I mention Harry Potter, a fan may recall the death of Dumbledore, but a person who did not read the books nor watched the movies will not produce that association.
LLMs literature misinterprets context to be "words", while I see context as phenomena and their interactions with history. It may be from memory or about one's beliefs, about abstractions (like in math) or imaginary (like Harry Potter) or hypothetical (like multiple worlds). I agree that poets do important work with respect to language, maybe I will try to address that in the future but at the moment I feel that my ideas need to be stated as clearly as possible. Any help is welcome. Your critical eye is so much appreciated. Thank you! Exactly what I and my theory need.
Alexander, thank you for engaging so extensively with the critique. You are generous in tone and clear about your intent, which is to develop an original theory of intelligence and language. That said, your reply confirms rather than resolves the problems raised earlier. What you are presenting is not yet a theory in any disciplinary sense, but rather a metaphorical framework that borrows freely from established domains without fully understanding or integrating them.
Let me try and respond to your points in turn.
1. "I have my own explanatory theory of intelligence..."
You begin with a statement that immediately requires scrutiny. The claim to possess a theory of intelligence carries significant weight. The term "theory" is not a placeholder for "framework I find intuitive." A theory, particularly in cognitive science or linguistics, must meet several conditions:
* Coherence: internally consistent definitions and operations
* Predictive power: ability to generate testable hypotheses
* Falsifiability: conditions under which it could be shown to be incorrect
* Integration: relation to existing research and empirical data
At present, your proposed "selection algorithm" that compares properties and selects options based on constraints is reminiscent of basic decision-making models in AI and cognitive science. Yet you provide no formalisation, no references to related models (e.g. ACT-R, Bayesian approaches, constraint-based grammars), and no empirical validation.
Calling it your own theory of intelligence does not grant it explanatory authority. It must earn that by engaging the existing literature and by making testable claims.
2. Objects, Properties, and Your Account of Language
Your theory of language is based on a reductionist schema: objects have properties; adjectives represent properties; nouns are sets of properties; actions change properties. While superficially attractive, this view is both under-theorised and overgeneralised.
* Properties as axes: You claim a property is an axis with unique semantics divided into ranges. This sounds like a rephrasing of feature-based semantics or dimensional models of meaning, such as those found in prototype theory or distributional semantics. Are you familiar with Rosch's categorisation work or Gärdenfors’ Conceptual Spaces? Without this context, your terminology remains disconnected from ongoing research.
* Concepts as ranges: The idea that concepts emerge from range boundaries is broadly compatible with theories in cognitive linguistics, but again, without citation, elaboration, or critical engagement, it floats as an assertion rather than an insight.
* "Actions change properties": This collapses a vast number of verb types, aspectual distinctions, and syntactic structures into a single relational schema. Vendler's classification of verb types, lexical aspect, and argument structure in syntax all offer much more nuanced and empirically supported accounts.
To claim explanatory power, a theory must do more than redescribe linguistic categories using new metaphors.
3. Your Dismissal of Linguistic Distinctions
Several responses in your reply point to a resistance to disciplinary concepts that appears more rhetorical than justified.
* "Exophoric vs endophoric deixis is artificial": This distinction is not a conceptual gimmick. It captures a crucial difference in reference resolution. Endophoric reference points within the text, while exophoric reference relies on situational context. Disregarding this collapses discourse analysis and reference pragmatics into a singular intuition about context.
* "Syntax is just convention": Reducing syntax to social convention ignores decades of empirical work showing that syntactic constraints are not only cross-linguistically patterned but also cognitively processed with remarkable consistency. It also ignores work on universals, processing models, and syntactic priming. Rules may have exceptions, but the presence of exception does not abolish the rule.
* "Compositionality is stacking filters": You reject the compositionality principle without understanding what it seeks to explain. Compositionality in semantics is not about meaning per se, but about systematicity—how complex meanings depend on the meanings of parts and their arrangement. Replacing this with a vague idea of filters again trades one metaphor for another without solving anything.
4. Empirical Claims without Evidence
You write: "Do I have to? I thought it is a well-established fact..." Yes, you do have to! If you are proposing a new theory, especially one that crosses disciplinary boundaries, you are responsible for grounding it in existing knowledge or showing where and why that knowledge is insufficient. Simply referring to a child development website is not adequate in a scholarly exchange.
If your theory involves language acquisition, then cite Tomasello, Clark, Gleitman, or Karmiloff-Smith. If it involves the relationship between perception and language, cite Jackendoff or Barsalou. Otherwise, it remains anecdotal.
5. On Names, Reference, and Rigidity
You attempt to counter Kripke’s theory of rigid designation by pointing out that many people share the name “Messi” or “Aristotle.” But this is a misunderstanding of Kripke’s point. Kripke’s theory is not about uniqueness, but about how a name refers, namely through a causal-historical chain. That some names are ambiguous in use is a problem of disambiguation, not of reference rigidity.
Your counterexample does not undermine the theory. It just misrepresents it.
6. Final Points: Context, Poetry, and Claims to Innovation
You end by distinguishing your own theory of context from that of LLMs and mention that poets do important work with language. These gestures are not unwelcome, but they remain vague. Your notion of context as "phenomena and their interactions with history" is interesting, but again, lacks formal content. Are you proposing an event-based ontology? A mental model framework? Are you aware of situation semantics or discourse representation theory?
More importantly, if you genuinely want to develop a new theory, you must accept the burdens that come with that ambition:
* You need to define your terms clearly and consistently.
* You need to show how your theory differs from existing ones, and why that matters.
* You need to show what your theory explains that others cannot.
* You need to be able to show where it can be tested, challenged, or extended.
Otherwise, what you have is not yet a theory. It is a set of personal intuitions expressed in idiosyncratic language, with metaphor standing in for argument.
Thank you for these comments. Now I know how I should present my ideas. I am about to write another post or two. I hope you will agree to take a look and comment. I find your attitude and expertise extremely helpful.
I have a new post, inspired by your comments. Will you please take a look? https://alexandernaumenko.substack.com/p/intelligence-and-language