Verbal symbols are spoken or written words that stand for something other than themselves, and they are the foundation of every human language on Earth. What makes them remarkable is not just that they carry meaning, but that the connection between a word and what it represents is almost entirely learned rather than built into the sound itself. The word “tree” sounds nothing like a tree. Yet every English speaker who hears it instantly conjures branches and leaves. This capacity to map arbitrary sounds onto meanings, share those mappings across a community, and recombine them into limitless new sentences is one of the defining features of the human mind, and research from neuroscience, developmental psychology, and comparative animal studies continues to reveal just how deeply that capacity is woven into human biology and culture.
What Makes a Sound a Symbol
A verbal symbol works because a community agrees, implicitly and through long use, that a particular sound pattern refers to a particular concept. The Swiss linguist Ferdinand de Saussure formalized this idea in the early twentieth century by distinguishing between the sound pattern (which he called the signifier) and the concept it points to (the signified). The crucial insight was that the relationship between the two is arbitrary: there is no physical reason why the sounds in “dog” should mean a four-legged canine. French speakers use “chien,” Mandarin speakers use “gÇ’u,” and neither word is more naturally dog-like than the other. That arbitrariness is actually a strength, because it frees a language to assign any sound to any idea, creating an open-ended system capable of expressing anything its speakers need.
This arbitrariness separates verbal symbols from most other forms of communication in the natural world. A growl may directly signal aggression; a whimper may directly convey pain. Those signals are tied to the physical or emotional state producing them. A verbal symbol, by contrast, can refer to things that are absent, imaginary, or abstract. You can say “justice” or “next Tuesday” or “unicorn” with equal ease, none of which are present in the room when you speak. That ability to talk about what is not here, sometimes called displacement, depends entirely on the conventional, shared nature of verbal symbols.
When Sound and Meaning Are Not Entirely Arbitrary
The arbitrariness principle has a well-studied exception. Certain sounds seem to carry built-in associations with particular shapes, textures, or sizes, a phenomenon researchers call sound symbolism. The best-known demonstration is the bouba-kiki effect: when people are shown a blobby, rounded shape and a jagged, spiky shape and asked which one is “bouba” and which is “kiki,” the vast majority assign “bouba” to the round shape and “kiki” to the spiky one.
This preference is not just a quirk of English speakers. A large cross-cultural study spanning multiple languages and writing systems found that roughly 72% of participants matched round-sounding words with round shapes across the board, with a high probability that the effect is above chance level.
What drives the effect? Research suggests it traces back to physical acoustics. Low-frequency, continuous sounds map onto round shapes, while high-frequency, discontinuous sounds map onto angular ones. Experiments using non-speech stimuli, like pure tones and noise bursts, confirmed that these spectral and temporal properties predict shape associations even when no recognizable word is involved.1Scientific Reports. Resolving the bouba-kiki effect enigma by rooting iconic sound symbolism in physical properties of round and spiky objects The effect extends into social perception as well: people associate round-sounding names with cooperative or friendly individuals and sharp-sounding names with competitive ones.2PubMed. Group Bouba-Kiki effects: The interplay of social categorization, competition, and sound symbolism
Sound symbolism does not overturn the arbitrariness principle; the vast majority of words in any language have no detectable resemblance to what they mean. But it does suggest that verbal symbols sit on a spectrum. At one end are fully arbitrary words like “table.” At the other are onomatopoeia like “buzz” and “splash,” where the sound imitates the referent directly. Sound symbolism occupies the middle ground, providing a faint, statistical tilt that may have helped early humans begin linking sounds to meanings before full-blown arbitrary convention took over.
How Children Crack the Code
Babies are not born knowing that sounds are symbols. They have to figure out that the noises adults make refer to things in the world, and the speed at which they do so is one of the most impressive feats in human development. By about 12 months, most infants understand a handful of words. By age two, they are adding several new words a day. By six, they know thousands.
A key part of this process is what researchers call fast-mapping: the ability to pick up the approximate meaning of a new word from just one or two exposures. Children do not need extensive drilling. They hear a word used in context, form a rough placeholder meaning, and then refine it over time through repeated encounters.3PubMed Central. Fast-mapping placeholders: Using words to talk about kinds That initial placeholder might be too broad, which is why a toddler might call every four-legged animal “doggy” for a while, or too narrow, refusing to apply the word “cup” to anything other than one specific cup. These overextensions and underextensions gradually resolve as the child accumulates more examples.
What matters most in those early encounters is not the form of the symbol but the social context in which it appears. Experiments with infants as young as 13 months showed that babies learned both words and non-word sounds (like a whistle) when an adult used them referentially, looking at and pointing to an object while producing the sound. When the same sounds were produced without that social-referential context, infants failed to learn either one.4PubMed. The role of social-referential context in verbal and nonverbal symbol learning The takeaway is that joint attention between caregiver and child provides the scaffolding that turns meaningless sounds into verbal symbols.
Caregivers also seem to adapt their communication style to help bridge the gap between sound and meaning. When talking to younger children about unfamiliar objects, especially objects that are not physically present, parents use more onomatopoeia and more iconic gestures, as if leveraging the non-arbitrary edges of language to bootstrap understanding of the arbitrary core.5Child Development. Language development beyond the here-and-now: Iconicity and displacement in child-directed communication As children get older and their vocabulary deepens, caregivers dial back these supports.
The Brain’s Symbol-Processing Network
Understanding a spoken word requires your brain to do at least two things almost simultaneously: recognize the sound pattern and retrieve the meaning attached to it. Neuroimaging research has revealed that these two jobs rely on distinct but interacting pathways.
A dorsal pathway runs from the auditory cortex upward through the parietal lobe toward the premotor regions in the frontal lobe, linking what you hear to how you would produce the sound yourself. A ventral pathway runs forward along the temporal lobe toward the ventrolateral prefrontal cortex, connecting sound patterns to meaning.6PubMed Central. Ventral and dorsal pathways for language Brain imaging studies that asked participants to perform tasks ranging from repeating nonsense syllables to judging word meanings consistently confirmed this division: the dorsal stream handles form-to-articulation work, while the ventral stream handles form-to-meaning work.7PubMed Central. Revealing the dual streams of speech processing
The meaning side of this equation extends well beyond a single brain region. A meta-analysis of 120 neuroimaging studies identified a left-lateralized network of seven brain regions consistently activated during semantic processing, spanning from the inferior parietal lobe to areas in the temporal and frontal lobes and including medial prefrontal cortex.8PubMed Central. Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies Some of these regions are modality-specific: hearing the word “hammer” activates motor areas associated with grasping, while hearing “red” activates visual color-processing areas. Other regions, particularly in the temporal lobe and inferior parietal lobe, sit at the convergence of multiple sensory streams and build more abstract, supramodal representations that tie together everything you know about a concept regardless of which sense originally delivered the information.9PubMed Central. The neurobiology of semantic memory
Even when no external stimulus is present, asking someone to imagine the sound of a trumpet or picture the color of a lemon engages both sensory cortex and these transmodal convergence zones. The transmodal areas, especially the anterior middle temporal gyrus, responded equally strongly to visual and auditory concepts yet still encoded enough modality information to allow researchers to decode which kind of feature the person was imagining.10PubMed. Imagining Sounds and Images: Decoding the Contribution of Unimodal and Transmodal Brain Regions to Semantic Retrieval in the Absence of Meaningful Input In other words, the brain’s symbol-processing network does not just match sounds to dictionary definitions. It reconstructs a rich, multi-sensory experience tied to each word.
When Verbal Symbols Fall Apart
If the temporal lobes are where your brain stores and retrieves word meanings, damage to those areas should selectively destroy the ability to use verbal symbols while leaving other mental abilities relatively intact. That is precisely what happens in a condition called semantic dementia, a progressive neurodegenerative disorder that primarily attacks the anterior temporal lobes. People with semantic dementia gradually lose the meanings of words. They remain fluent; they can string grammatically correct sentences together. But they can no longer name objects or understand what common words refer to.11PubMed Central. Clinical Manifestations and Neural Basis of Semantic Dementia: Converging Evidences From Brain Imaging Studies
Early clinical descriptions documented patients who progressed from occasional word-finding difficulty to a near-complete dissolution of the semantic components of language, losing knowledge about a wide range of living and manufactured items.12Brain. Semantic Dementia: Progressive Fluent Aphasia with Temporal Lobe Atrophy Computational modeling of these patients has shown that a single underlying factor, the gradual loss of activation capacity in semantic memory, can account for nearly all of the variation in naming, comprehension, and repetition performance, explaining about 95% of variance at the individual patient level.13PubMed. Accounting for word production, comprehension, and repetition in semantic dementia, Alzheimer’s dementia, and mild cognitive impairment
Semantic dementia is distinct from other language disorders. In Broca’s aphasia, for instance, patients know what they want to say but cannot produce fluent speech; their symbols are intact but the motor-articulatory system is damaged. In semantic dementia, the production machinery works fine, but the symbols themselves have been emptied of content. A patient might look at a picture of a camel and call it “an animal, or something” because the specific meaning has dissolved while the general category has not yet been lost. The condition serves as a kind of natural experiment, demonstrating that verbal symbols are not a unitary thing in the brain. The sound pattern, the articulatory program, and the meaning are handled by different neural systems, and each can fail independently.
Evolutionary Roots of Verbal Symbols
How humans came to use verbal symbols at all is one of the oldest and most contentious questions in science. The archaeological record offers some clues. Engraved ochre and ostrich eggshell fragments from sites in South Africa, dating back roughly 100,000 years, show signs of becoming more salient, memorable, and reproducible over tens of thousands of years. But experiments using these engravings as stimuli found no clear indication that they served as denotational symbolic signs pointing to specific meanings the way words do. They appear to have functioned more as decorative or culturally transmitted patterns.14PubMed Central. The evolution of early symbolic behavior in Homo sapiens That means the leap to true verbal symbols, where a sound or mark stands for a specific idea by community agreement, may have happened without leaving physical traces for archaeologists to find.
A longstanding debate in language evolution research is whether verbal symbols originated in vocalizations or in gesture. Experiments in which people had to communicate meanings without using language found that gesture was consistently better adapted to meaning transfer than non-linguistic vocalization, both within and across cultures, and even when the gesturers were severely vision-impaired.15PubMed Central. Gesture is the primary modality for language creation Proponents of a gesture-first theory argue that the iconic, imitative qualities of gesture gave early humans a way to communicate about absent or novel things before fully arbitrary vocal symbols took hold.16PubMed Central. Language, gesture, skill: the co-evolutionary foundations of language
Others push back on the idea that one modality came first. An analysis of the evidence concluded that language likely evolved in oral-aural and gestural modalities together, with neither taking precedence.17PubMed Central. Reflections on the “gesture-first” hypothesis of language origins Under this view, early human communication was multimodal from the start, and the shift toward predominantly vocal symbols happened gradually as vocal control improved. One piece of biological evidence supporting the importance of vocal capacity is the gene FOXP2: a single point mutation in FOXP2 co-segregates with severe speech and language impairment in an affected family, and disruption of the gene in an unrelated individual produced a similar disorder, suggesting that two functional copies are required for normal spoken language development.18PubMed. Molecular evolution of FOXP2, a gene involved in speech and language
What Other Animals Can and Cannot Do
Other species communicate in sophisticated ways, but the gap between their communication and human verbal symbols is striking. Vervet monkeys, for example, produce distinct alarm calls for different predators, and other group members respond appropriately, looking up for eagles and scanning the ground for snakes. Research has also demonstrated that vervets interpret these calls in context: when an alarm call came from a speaker placed where the caller was known to be present, the monkeys showed lower vigilance and fewer startle responses than when the same call came from an unexpected location.19PubMed Central. Context-dependent alarm responses in wild vervet monkeys These calls refer to things in the world, which is a precondition for being a symbol. But they are limited to a small, fixed set of meanings tied to immediate situations, and vervets do not recombine them the way humans combine words into novel sentences.
Great apes trained in laboratory settings come closer. Chimpanzees and bonobos have learned to use visual symbols called lexigrams to request food, comment on events, and combine symbols to express simple semantic relations like agent-action or action-object. All three apes in one comparative study used ordering strategies for combining symbols, showing commonalities with the language of two-year-old children.20First Language. Semiotic combinations in Pan: A comparison of communication in a chimpanzee and two bonobos Yet the differences were large: combinations made up a much smaller proportion of ape production, single symbols dominated, and ape combinations rarely exceeded three elements. On the comprehension side, greater abilities have been reported: the bonobo Kanzi, for instance, could respond correctly to novel reversible sentences like “make the doggie bite the snake,” which requires tracking word order to determine who does what to whom.21PubMed Central. Evidence of Grammatical Knowledge in Apes: An Analysis of Kanzi’s Performance on Reversible Sentences
These findings suggest that the raw cognitive ingredients for verbal symbols, learning that an arbitrary signal maps to a concept, are not unique to humans. What appears to be unique is the explosive combinatorial capacity that lets us string symbols together into unlimited new meanings and the cultural infrastructure that transmits thousands of those mappings to every new generation without explicit teaching.
Do Your Words Shape What You See?
If verbal symbols are arbitrary labels draped over a pre-existing reality, they should not change how you perceive that reality. But decades of research on linguistic relativity, the idea sometimes loosely called the Sapir-Whorf hypothesis, suggests that verbal symbols do subtly influence perception and memory. The domain where this has been tested most rigorously is color. Different languages draw boundaries between color categories in different places: Russian has separate basic words for light blue (“goluboy”) and dark blue (“siniy”), while English uses one word, “blue,” for both.
Experiments have found that these linguistic boundaries affect how well people discriminate between colors. Cross-language differences in color discrimination appear to result from category-induced bias under uncertainty, where people’s memory for ambiguous colors is pulled toward the center of their language’s color category.22PLOS ONE. The Sapir-Whorf Hypothesis and Probabilistic Inference: Evidence from the Domain of Color The effect is not that speakers of different languages see different wavelengths of light. It is that when you have to remember or compare colors under conditions of uncertainty, the verbal labels your language provides act as an anchor, nudging your judgment toward the category center. The verbal symbol becomes part of the perceptual process.
This finding has practical implications beyond the color lab. Legal eyewitness descriptions, cross-cultural design work, and even medical imaging interpretation all involve situations where a verbal label might subtly reshape what a person remembers seeing. The influence is modest, not a wholesale rewriting of perception, but it is consistent and replicable, and it underscores the point that verbal symbols are not passive labels sitting on top of thought. They participate in it.
How Word Meanings Drift Over Time
Because the link between a verbal symbol and its meaning is conventional rather than natural, that link can shift. Words gain new meanings, lose old ones, and sometimes flip entirely. “Nice” once meant foolish; “awful” once meant awe-inspiring. This kind of semantic change is not random. Research tracking meaning shifts over the past 200 years found that words acquired later in life and those that are harder to process are more likely to undergo semantic change.23PubMed Central. Diachronic semantic change in language is constrained by how people use and learn language Early-learned, easy-to-process words like “mother” or “water” tend to be semantically stable, which makes intuitive sense: the more deeply a word is entrenched in the community’s shared vocabulary, the harder it is to dislodge its meaning.
Semantic change also carries a measurable cognitive cost for people who lived through the shift. Middle-aged adults who grew up with one set of meanings responded more slowly to words whose meanings had changed substantially between 1970 and 2000, while younger adults who learned the updated meanings from the start showed no such slowdown.24PubMed Central. Diachronic semantic change in language is constrained by how people use and learn language The old mapping between sound and meaning does not simply get overwritten; it lingers, creating interference. This is a vivid demonstration that verbal symbols are not stored like dictionary entries that can be edited cleanly. They are embedded in a web of associations laid down through years of use.
Can Machines Really Use Verbal Symbols?
Large language models like GPT-4 process text that is, at bottom, a sequence of verbal symbols encoded as numbers. They learn statistical patterns among those symbols and produce remarkably fluent output. But whether they are actually using verbal symbols in the way humans do is a genuinely open question in cognitive science. The central issue is grounding: human verbal symbols derive their meaning partly from sensory and motor experience with the world. You know what “cold” means not just because of its relationships to other words but because you have felt cold. A language model trained only on text has never felt anything.
Some researchers have argued that this may not matter as much as it seems. If the structure of relationships among concepts learned by an ungrounded language model turns out to be the same structure that exists in grounded human representations, then in principle, any computation carried out in the grounded space could be carried out in the ungrounded space with the same results.25PubMed Central. Symbols and grounding in large language models Others counter that evidence from human cognition shows semantic memory is partially grounded in sensorimotor systems and depends on language-specific learning in ways that purely text-based models cannot replicate.26PubMed Central. Symbol ungrounding: what the successes (and failures) of large language models reveal about human cognition
An alternative perspective frames the problem differently. Rather than asking whether a machine’s symbols are grounded in perceptual experience, some theorists propose that a symbol is any structural coupling between an agent’s sensorimotor activations and its environment.27Cognitive Systems Research. The physical symbol grounding problem Under that definition, a robot that learns the meaning of “hot” by repeatedly touching surfaces and measuring temperature has grounded symbols, while a text-only model does not, no matter how sophisticated its word associations become. The debate is unresolved, but it highlights something important about verbal symbols in general: for humans, at least, meaning is not just a pattern in a network. It is tethered to lived experience in ways that the brain’s sensory and motor systems help maintain. Whether that tethering is essential to meaning, or merely the particular way human brains happen to implement it, is one of the more consequential questions in modern cognitive science.