Two key features of human cognition are language and conceptual categorization [1]. Developmental studies have shown that during the first years of life, verbal (spoken words)—as opposed to non-verbal meaningless—sounds facilitate conceptual categorization [2, 3]. Conceptual categorization implies constructing an abstract mental representation of a category by grouping different exemplars or objects into clusters based on shared features, such as perceptual, functional, taxonomic or thematic [4]. Consequently, upon hearing the label of an object (e.g. ‘dog’), all category related objects (e.g. exemplars of dogs, and/or associated animals (e.g. cat, sheep) are activated, which in turn, results in faster visual object recognition [5, 6]. Meaningful auditory information, however, originates not only from spoken words (e.g. “dog”), but also from environmental associated sounds (e.g. bark). Although words and associated sounds are both informative and semantically related to their referents (e.g. dog), they fundamentally differ from one another [7]. While associated sounds are based on causal relationships being strictly related to their generating source, words are arbitrarily linked to their referents, have phonological forms that are reproduced by a person, carry an informative intent, and are used to label objects or name a category to which these objects belong [8, 3].

Though both words and associated sounds carry semantic knowledge, research on language processing and object recognition has focused mostly on spoken words, and much less on associated sounds. It remains unclear whether during language development, infants process words and associated sounds similarly, or whether the former has an advantage over the latter. Uncovering what effects words have on developing conceptual representations can contribute to a better understanding of the relation between language and cognition. An important question is whether words and associated sounds activate conceptual representations differently, and consequently, whether recognizing visual information can be modulated by a preceding auditory cue (e.g. word versus associated sounds). In adults, a few studies that investigated the semantic organization of words and associated sounds have shown that compared to words, associated sounds enhanced visual object detection (judging whether an object was present on the visual display) [9], especially when the stimulus onset asynchrony (SOA; time from the beginning of the auditory stimulus to the appearance of the image) was short (e.g. 350 ms; 10). Chen & Spence [10]suggested that words access their meanings via lexical representations, whereas associated sounds access faster and more directly their meaning. The findings of a more recent study using the visual world paradigm (VWP), revealed that participants looked faster at the target (e.g. puppy) and longer at its competitor (e.g. bone) compared to other two distractors (e.g. candle and daffodil), suggesting similar graded effects for both associated sounds and words [11]. In addition, the results showed that this graded pattern was more pronounced in the associated sounds condition [11].

On the other hand, when investigating the activation of conceptual representations during object recognition, Lupyan and Thompson-Schill [8]found that words (e.g. ‘dog’) activate conceptual representations more effectively than associated sounds do (e.g. dog bark). In a series of visual identification tasks, adults systematically reacted faster to a target image when primed by a word compared to an associated sound. In another VWP study [6], the results also revealed that sound primes led participants to look more at one category exemplar (e.g. robin) compared to the three others (e.g. three different exemplars of bird), most likely the source of the sound, whereas in response to word primes, participants looked equally at the four images. These findings suggest that while words are somehow detached from the perceptual information, sounds are tightly linked to the perceptual details of the generating source [6]. Furthermore, although both words and sounds yielded similar N400 response (an event-related brain potential (ERP) known to reflect semantic processing; [12]), words elicited earlier and larger P1 ERP response, which is related to perceptual categorization [5].

Altogether, Lupyan and colleagues’ work contrasts previous hypotheses that words and sounds are processed similarly [13], or that sounds access their meanings faster than words [10]. They demonstrate that although both words and sounds activate conceptual representations, the representations activated by words are enhanced, facilitating the match to the category exemplars. Unlike associated sounds, labels are abstract symbols “standing for” and referring to objects. They are used by humans to communicate and convey abstract information that is not strictly linked to the ‘here and now’ of an object, whereas associated sounds are mere features of objects [8, 14]. According to Edmiston and Lupyan [6], sounds act as “motivated” cues, and are idiosyncratically linked to their referents, whereas words are decontextualized “unmotivated” cues, and activate conceptual categories abstractly. By taking this stance, we were interested in determining whether differences in processing words and associated sounds occur early in language development. This will contribute to a better understanding of how we form categories, and to the theoretical account stating that ‘words refer to’ [3], rather than being merely ‘associated to’, objects (3, for a review). According to this account, words enable more abstract conceptual representations and are not directly linked to the context or event, therefore enhancing object recognition.

To the best of our knowledge, there are only two developmental studies that investigated whether young children process known words and sounds similarly [15, 16]. In Cummings et al. [15] study, 15-, 20-, and 25-month-old toddlers participated in a looking-while-listening task, during which they viewed pairs of images (e.g. dog–piano) and heard either associated sounds (e.g. dog barking or piano playing) or words. The results showed that across ages, infants were equally fast at recognizing the target object in both word and associated sound conditions. Faster object recognition preceded by words was correlated with infants’ productive skills. In their recent ERP study, Hendrickson et al. [16]investigated the semantic organization of words and associated sounds in the developing brain of 20-month-olds, including three control conditions. Toddlers viewed the target images (e.g. dog) while hearing matching words (e.g. “dog”) or associated sounds (e.g. barking), within-category violations (“cat” or meowing), and between-category violations (e.g. “pen” or scribbling). The ERP results showed that 20-months-olds exhibit different patterns of brain activation in response to words and associated sounds. While between-category violations (e.g. dog–“pen” or scribbling) elicited similar ERP responses across words and associated sounds, within-category violations (e.g. dog–“cat”) for words elicited earlier and greater negativity than for associated sounds (e.g. dog–meowing), suggesting that young children exhibit greater sensitivity to the relationship between words than that of associated sounds in the semantic system.

The current study aimed at extending these findings, exploring whether words have an advantage over associated sounds in activating conceptual representations in infancy as they have in adulthood. If early in development, infants, like adults, interpret words as abstract, “unmotivated” and arbitrary symbols, and sounds as “motivated” and idiosyncratic cues, the visual object recognition should be modulated by the preceding auditory information. Thus, like in Lupyan and Thompson-Schill [8]study, activation of conceptual representations would be more efficient when target objects are cued by words than by associated sounds. Alternatively, if words and sounds activate conceptual representations similarly, object recognition will not be modulated by its preceding cue. Our first goal was therefore to replicate the study of Lupyan and Thompson-Schill [8] with adults, by using a similar behavioural visual identification task (Experiment 1A). We then conducted an object recognition task with adults (Experiment 1B), measuring eye movements. This eye tracking task was also suitable for infants at 9- (Experiment 2A), 12- (Experiment 2B), and 18 months of age (Experiment 2C). We predicted that adults will react faster to the target image (e.g. cow) when preceded by a spoken word (e.g. “cow”) compared to meaningful associated sounds (e.g. cow mooing). This should also be reflected by faster and longer looking behaviour to the target image (e.g. cow) compared to a distractor (e.g. train). Similarly, we predicted that infants will look faster and longer at the target object when preceded by word compared to associated sound primes, indicating that the advantage of words emerges early during language development.

In this study, we aimed to determine whether during language development, words are more effective than associated sounds at activating conceptual representations. We conducted one behavioural visual identification and one eye tracker object recognition experiment with adults to replicate Lupyan and colleagues’ [8]findings. We then adapted the object recognition task so that it was suitable for infants. Our successful replication revealed that adults identified and recognized faster the target object when preceded by its label compared to its associated sound, supporting further the theory that although both words and associated sounds activate conceptual representations, words have the advantage of being more efficient in activating the visual representation of an object. In Lupyan and Thompson-Schill’s study [8], the word advantage was also evident in the accuracy measure, whereas in our study, participants were equally accurate across words and associated sounds conditions. The eye movement measures, however, strengthened the RTs findings and yielded a similar word advantage.

It is possible that upon hearing the word “dog” for instance, all dog features including their generic visual appearance are activated, accelerating the reaction times, while hearing dog barking might require increased verification time to create the direct link between the source of the sound and the image. This is unlikely, however, as even with a longer average sound duration (relative to word duration) providing participants with additional processing time, RTs and looking times were faster in the word compared to the associated sound condition. Thus, unlike associated sounds, words enhance the abstraction of conceptual categories leading to faster activation of the category representations: words are “special” because they enable activation of conceptual representations in a more categorical way [3, 8].

However, this phenomenon did not emerge in our experiments with infants at 9-, 12-, or 18-months. Nine- and 12-month-olds did not display any differences between words and associated sounds conditions. Moreover, the distinction between the target and distractor object in either condition was not reliable. Consequently, it can be assumed that at these ages, infants process both auditory stimuli differently, but our empirical paradigm was not sensitive enough to detect these nuances. Unexpectedly, although 18-month-old infants were equally fast at recognizing the target object in both conditions, we observed a shift towards larger fixations and longer looking time at the target when preceded by the associated sound compared to the word. Our findings suggest that infants allocated greater attention to the target image in the associated sound compared to the word condition. This result must be taken with caution and needs to be considered carefully. Though it matches Hendrickson et al. [16]suggestion that associated sounds require longer time to process the semantic match between the visual object and the generated sound, it contradicts the results of Cummings and colleagues [15], whereby object recognition was similar in the words and associated sounds condition. Toom and Kukona’s [11]VWP study with adults, found greater looking times and semantic activation of the competitors in the associated sounds relative to the words conditions. Just like in our study, however, this source of differences requires further replication and investigation.

Alternatively, developmental studies commonly claim that words have a special status and are preferred over other non-linguistic sounds because of their referential nature [3]. In this account, words enhance categorization and learning because unlike other sounds, words refer to object kinds. Therefore, it could be that during infancy, words activate conceptual representations more efficiently than associated sounds, but another methodology would be more sensitive to these differences. We also want to note that, although the ISI was kept constant (1000 ms), for both adults and infants, the average duration of words was shorter compared to sounds, and participants had more time to process sounds over words. As Exp. 1B indicates, this had no effect on adult results but might have had an effect on infant results.

Another possibility is that words become more effective at activating conceptual representations, but their advantage over associated sounds emerges at later stages of language development. The only two studies to date that investigated the processing of words and associated sounds tackled the question differentially. Cummings et al. [15]studied the speed of word-object recognition and its correlation with chronological age and infants’ productive skills, while Hendrickson et al. [16]investigated the organization of words and associated sounds in the semantic memory. Here, we were interested in studying whether visual object recognition is modulated by the preceding auditory cue to determine whether words have a ‘special’ status compared to associated sounds.

Our study demonstrates that by 18 months, infants process words and associated sounds differently, possibly allocating more attention to target objects when cues by associated sounds relative to words. The question of whether and when infants reach the pattern of results we observed in adults remains open: a different experimental methodology or different ages might yield the initially expected results.