Being unable to construct a sentence is not the same as being unable to understand a structured problem. That distinction gives the debate about language and thought one of its most consequential tests. If complex reasoning always required intact grammatical processing, damage to that processing should remove the reasoning too.
In 2005, Rosemary Varley and colleagues reported mathematical testing with three men who had severe agrammatic aphasia, a language impairment involving profound difficulty with grammar. The participants nevertheless retained substantial calculation abilities, including operations sensitive to order and brackets. “Agrammatic but Numerate”.
These were individuals participating in specific research tasks, not symbols of a universal claim about every person with aphasia. Their performance matters because it puts pressure on a claim of necessity. A process that can survive a serious impairment of another process cannot simply be assumed to depend on it at every moment.
Structure need not be sentence structure
Consider an ordinary arithmetic example, separate from the study's reported items. The expressions 18 − 7 and 7 − 18 use identical symbols in a different order. They produce 11 and −11. Order is part of the operation, not decoration.
Brackets introduce another kind of structure. 4 × (3 + 2) gives 20, whereas (4 × 3) + 2 gives 14. A solver must group the parts correctly. Going from left to right without attending to the grouping will not reliably give the answer.
The research tested analogous kinds of structural demands across language and mathematics. The participants' success with mathematical structure contrasted with their difficulty processing sentence structure. The authors interpreted the dissociation as evidence that an established adult system can sustain calculation without intact linguistic grammar. Varley and colleagues, results and discussion.
The arithmetic examples make the distinction visible without turning the reader into a research participant. They show what it means for an operation to depend on a relationship among symbols. “Structured” does not mean “encoded as a sentence.” Nor does every symbolic system have to use the same rules simply because both can be described as having structure.
Learning a capacity and using it are different questions
The adult participants had acquired language and mathematics before their impairments. Their preserved abilities therefore do not establish that language played no part in learning those abilities. The study explicitly distinguishes independence in an established system from the question of how that system develops. Original discussion.
A hypothetical analogy helps clarify the logic. A person can use a procedure after the teacher who explained it has left the room. That does not show the explanation was irrelevant to learning. Equally, receiving an explanation does not mean mentally repeating it is necessary each time the procedure is used.
The analogy is only about the difference between acquisition and current performance. It does not explain the brain mechanisms of aphasia. The research question remains narrower: which resources are required now, and which may have helped establish the capacity earlier?
This distinction prevents an argument from swinging between two extremes. Evidence that language can help learning is not evidence that all later thought is linguistic. Evidence that a learned ability survives grammatical impairment is not evidence that communication contributes nothing to development.
Some quantity discrimination precedes symbolic counting
Fei Xu and Elizabeth Spelke's 2000 study reported that six-month-old infants discriminated sets of eight versus sixteen objects, but not eight versus twelve, under controls for other variables. The authors interpreted the ratio-dependent pattern as evidence for approximate numerical representation before language and symbolic counting. “Large Number Discrimination in 6-Month-Old Infants,” original abstract.
This is an early capacity with a particular limit. It is not proof that infants perform exact arithmetic, know the meanings of number words, or understand the mathematical expressions above. Approximate discrimination and exact calculation are distinct operations.
That difference is useful rather than disappointing. A theory of development has to account for early capacities and for what later learning adds. Treating a young infant as either completely lacking number or already possessing adult mathematics leaves too little room for the actual pattern.
Communication can matter without supplying every thought
In a 2024 perspective, Evelina Fedorenko, Steven Piantadosi, and Edward Gibson argued that language in modern humans primarily serves communication and is not a prerequisite for complex thought. They also emphasized its power in transmitting cultural knowledge. That is the authors' synthesis of evidence, not a verdict that every possible cognitive role of language has been eliminated. “Language Is Primarily a Tool for Communication Rather Than Thought,” original abstract.
Communication's importance survives the narrower claim. Instructions can teach a method, a conversation can expose an error, and a written explanation can make another person's reasoning available for inspection. None requires the reader to believe that an unsaid thought is an impossible thought.
The distinction is especially important when listening to someone whose fluent expression is limited. Difficulty with a sentence gives evidence about that performance. It should not be used as an effortless measure of everything the person understands. At the same time, a documented preserved ability should not become a promise about another individual's capacities; people and impairments vary.
The boundary belongs to the evidence
A language-related difference in one task does not establish the outer limit of a community's intelligence. A preserved mathematical operation does not settle every question about language and reasoning. Each result addresses a particular proposed dependence.
The strongest claim—thought cannot proceed without the relevant words or grammar—has to meet the examples where meaningful performance does proceed. Those examples make room for a richer account: linguistic resources can help organize, teach, communicate, or support an operation without being the material from which every thought is made.
A sentence can fail while a calculation succeeds. Keeping both facts in view is more informative than letting fluent speech stand in for the whole mind.
Loading comments…