Dog-Directed Speech: Why We Talk to Dogs Like Babies

1. What Dog-Directed Speech Is

1.1 A Recognizable Way of Talking

Dog-directed speech is the term researchers use for the particular way people talk when they address a dog. It is a register: a set of changes in voice, vocabulary and sentence structure that speakers adopt when the listener changes.

The first systematic description came from developmental psychology. Hirsh-Pasek and Treiman (1982) analyzed how four owners talked to their dogs and compared that language with published descriptions of "motherese", the speech style adults use with young children. They called the dog version "doggerel" and found its structural properties strikingly similar to motherese, while noting that certain differences might arise in functional and social areas. Four dyads are a very small basis, but the paper set the question later research took up: what is this style actually for?

Later observational work described what people say. In a study of 23 people who each played with a familiar and an unfamiliar dog, an interaction contained about 208 words on average, and just eight words accounted for more than half of everything that was said (Mitchell & Edmonson, 1999). The most frequent phrases were "come on" and "come here". Imperatives were the most common sentence type, followed by attention-getting devices, statements about the dog and questions. The authors concluded that this repetitive talk had some conversational aspects but mostly attempted to control the dog.

1.2 The Acoustic Profile

Acoustic studies in English, French, Hungarian and German-speaking samples agree on the core features. Compared with speech to adults, dog-directed speech has a higher mean pitch, in most studies more pitch variation, shorter and more repetitive utterances, and many imperatives and attention-getters. In the earlier literature, as summarized in one acoustic study, infant-directed and dog-directed speech share these points: both have higher overall pitch and pitch variation, a lower mean length of utterance and more repetition than adult-directed talk (Mitchell, 2001; Gergely et al., 2017).

One production study recorded 30 women aged 17 to 55 (Ben-Aderet et al., 2017). Each woman spoke a fixed script ("Hi! Hello cutie! Who's a good boy? Come here!") while looking at photographs of a puppy, an adult dog and an old dog, and spoke it once more to a researcher. Eleven of 17 acoustic features differed between dog-directed and normal speech. Dog-directed speech was higher in pitch, more variable in pitch and higher in harmonicity. Compared with normal speech, mean pitch rose by about 21% for the puppy picture, 11% for the adult dog and 13% for the old dog. The authors noted that speaking to photographs reduces ecological validity.

A later study with live experimenters measured the same core features in speech recorded for dogs: dog-directed prosody had a higher pitch, greater pitch modulation and a slower word rate than adult-directed prosody (Benjamin & Slocombe, 2018). The higher pitch is the most robust finding across studies. How dogs perceive such sounds in the first place is covered elsewhere (how hearing in dogs works).

1.3 Pitch Range Depends on the Situation

Not every feature holds in every setting. In a study of 53 owners recorded during an adapted version of the Strange Situation procedure, speech to the dog was higher in pitch than read speech and conversation, and pitch was highest when owners called their dogs (Lesch et al., 2019). But the pitch range was narrower than in reading, not wider. The authors wrote that their results "fail to support the previously described broader pitch range" and suggested that pitch may carry affect while range may serve attention-getting. Since the recordings came from a stress test in one laboratory, this shows that context can change the register, not that earlier findings were wrong.

The influence of context was tested directly in 34 women talking to their own dogs in four situations: before a brief separation, after reunion, during play and while giving commands (Jeannin et al., 2017b). Before separation, the women spoke with low pitch, few modulations, high intensity variation and very few affective sentences. At reunion the pitch was very high, imperatives were rare and affectionate nicknames were frequent. Play produced mainly questions and attention-getters, commands mainly imperatives combined with attention-getters. There is no single dog-directed voice; "baby talk" is closest to the reunion and play pattern.

1.4 Dogs, Infants and Adults Compared

Two studies recorded the same speakers addressing an adult, an infant and an animal. In the first, 12 Australian mothers spoke to their six-month-old infant, to their pet (five cats and seven dogs) and to another adult (Burnham et al., 2002). Pitch in pet-directed and infant-directed speech was statistically equivalent, and both were higher than in adult-directed speech. Rated affect, judged from speech filtered so that the words could not be understood, was highest for infants, intermediate for pets and lowest for adults.

The second study recorded 39 Hungarian parents (21 mothers, 18 fathers) from 21 families, each with a child under 30 months and an adult dog, in four speech situations (Gergely et al., 2017). Adult-directed speech differed from both infant- and dog-directed speech regardless of speaker sex and situation. In the free situations, mean pitch was higher in dog-directed speech than in infant-directed speech. Pitch range was similar in both and wider than in adult-directed speech, except during task solving, where dog-directed speech did not differ from adult-directed speech. The authors noted that the higher pitch toward dogs could be specific to Hungarian, a fair caveat given that the Australian study found no difference.

1.5 The Missing Vowels

The clearest difference between the two registers concerns vowels. When adults talk to infants, they stretch the vowel space: the vowels become acoustically more extreme and more distinct from one another. This was shown across English, Russian and Swedish, and the authors argued that it provides infants with well-specified information about the units of their language (Kuhl et al., 1997). This feature is called vowel hyperarticulation.

Dog-directed speech lacks it. In the Australian study, the vowel triangle, a standard measure of how far apart the corner vowels are, was larger in infant-directed than in adult-directed speech, but pet-directed speech did not differ from adult-directed speech (Burnham et al., 2002). The authors described vowel hyperarticulation as a "didactic device" and summarized: "Mothers exaggerate their vowels for their infants but not for their pets." A follow-up with 11 adults who each spoke to an adult, a dog, a parrot and an infant found the largest vowel space for infants, a smaller one for the parrot and the smallest for adults and dogs (Xu et al., 2013). The Hungarian data show the same pattern: hyperarticulation in infant-directed speech, but none in dog-directed speech, which matched adult-directed speech (Gergely et al., 2017).

The authors interpret this pattern as a sign that people adjust vowels to the listener's actual or expected linguistic competence. If this reading is correct, the parts of dog-directed speech that people share with infant-directed speech are the emotional and attention-related parts (pitch, melody, repetition), and the part thought to support language learning is the part people leave out. Conference reports of hyperarticulated vowels toward pets exist, but the peer-reviewed evidence consistently finds none toward dogs.

2. Why Humans Do It

2.1 Four Hypotheses

No study has tested directly why people use dog-directed speech. The following are authors' interpretations of production and playback data: hypotheses, not findings, and not mutually exclusive.

Hypothesis 1: the register follows the listener's social responsiveness. Hirsh-Pasek and Treiman (1982) concluded that motherese may not be elicited by a child's linguistic or cognitive level, since dogs receive a similar style. They suggested that "the social responsiveness of the listener may be sufficient".

Hypothesis 2: the register is part of caregiving behavior extended to another species. Prato-Previde et al. (2006) interpreted owners' behavior toward their dogs as "interspecific parental behaviour" and offered an evolutionary interpretation of the finding that women talked more to their dogs. Benjamin and Slocombe (2018) raised the related question of whether dog-directed speech is an overgeneralization of infant-directed speech, at least within Western cultures. Both are interpretations that the studies did not test.

Hypothesis 3: the register is a spontaneous way to ease interaction with non-verbal listeners. Ben-Aderet et al. (2017) found that puppies, but not adult dogs, responded more strongly to dog-directed speech (see section 3). Because people use it with adult dogs too, the authors suggested that, with older dogs, it "may mainly be a spontaneous attempt to facilitate interactions with non-verbal listeners".

Hypothesis 4: the register serves control and attention. The observational data on play describe talk that is dominated by imperatives and attention-getting devices (Mitchell & Edmonson, 1999), and the stress-test study proposed that pitch range may serve to get attention (Lesch et al., 2019).

2.2 Who Speaks: Sex, Familiarity and Personality

Two small observational studies looked at differences between speakers. In a study of 8 men and 15 women playing with their own dog and with another person's dog, women used the baby-talk register more than men, and both sexes used it more with the unfamiliar dog (Mitchell, 2004). With the unfamiliar dog, speakers also used more praise, a more diverse vocabulary and longer utterances. Overall, the author emphasized that men and women used similar forms of talk.

In an Italian study of 25 owners (10 men, 15 women) observed in an adapted Strange Situation, women talked more, started talking sooner, and their speech resembled infant-directed speech more closely (Prato-Previde et al., 2006). There were no clear sex differences in affiliative behavior, play or reported attachment. The Strange Situation is a procedure borrowed from infant research to study how dogs respond to separation from and reunion with their owner (what attachment styles in dogs mean).

In the Austrian stress-test study, owner sex and the personality trait openness affected pitch range and voice jitter, whereas the owner's caregiving and the dog's attachment classification did not (Lesch et al., 2019). The register varies with the speaker, but the samples are small and the methods differ (what is known about sex differences in dogs; how owner personality relates to dog behavior).

3. How Dogs Respond

3.1 The Logic of Playback Studies

To test the register itself, researchers have to separate the voice from everything else a speaker does, such as leaning forward or smiling. Most studies therefore play recordings through loudspeakers while no person is visible, or ask owners to sit with their back to the dog and wear headphones. This prevents the dog from responding to unintended body signals (why unintended cues matter in dog studies). The price is an artificial situation: a voice without a body.

The outcome measures are indirect: looking time, orienting, approach or composite scores, all proxies for attention or interest. None of them measures understanding, learning or the relationship between dog and person.

3.2 Puppies Respond, Adult Dogs Do Not: Ben-Aderet et al.

In the first playback experiment of Ben-Aderet et al. (2017), the recordings of the 30 women were played to 10 shelter puppies aged two to five months and 10 shelter adult dogs aged 13 to 48 months in New York, plus 10 French family dogs who heard a French version of the script, so that 20 adult dogs were tested in total. The researchers combined 11 behaviors into one composite score. Nine of the 10 puppies responded more strongly to puppy-directed than to adult-directed speech. The adult dogs did not differ: 11 of 20 responded more to the dog-directed version, which is close to chance. Whether the dogs came from a shelter or a family made no difference.

A second experiment with 10 other puppies aged three to eight months compared speech recorded for the puppy picture with speech recorded for the adult dog picture and found no difference. The puppies' reactions were strongly influenced by mean pitch, whereas the adult dogs' reactions were not. With 10 puppies per experiment and only female voices, these results are a starting point rather than a firm answer.

3.3 Adult Dogs Respond, Puppies Do Not: Jeannin et al.

A French study published in the same year came to the opposite conclusion for adult dogs (Jeannin et al., 2017a). Nine women were recorded saying "On va se promener?" ("Shall we go for a walk?") to a real dog, a real infant aged three to four months and an adult. Seventy-one pet dogs were tested; 8 were excluded because they did not react, leaving 44 adult dogs (mean age 3.74 years, range 1 to 14.25) and 19 puppies (two to six months). Attention was measured as gaze duration toward the loudspeaker.

The adult dogs were significantly more attentive to pet-directed than to adult-directed speech, and their attention increased with the pitch of the women's voices. The puppies showed no significant difference between speech types, although they reacted more to all stimuli in general. Infant-directed and adult-directed speech did not differ significantly, and the dogs habituated quickly to repeated infant-directed speech. The authors wrote that their results "do not confirm" those of Ben-Aderet et al.

3.4 Why the Two Studies Disagree

Jeannin et al. (2017a) attributed the conflict to differences in protocol. In one study, speakers addressed photographs; in the other, a real dog. One study used a composite behavioral score; the other used gaze duration. The samples also differed: shelter puppies and a mixed group of adult dogs in one, pet dogs in the other, with 10 puppies against 19, and 20 adult dogs against 44.

Neither study settles the age question: the puppy samples are small, and the measures are not comparable. The disagreement is itself informative, since the result seems to depend on how the speech was produced and on what counts as a response. A general claim such as "puppies need baby talk" or "adult dogs ignore it" is not supported by either study. Developmental changes in how puppies respond to their environment are an obvious candidate explanation, but the studies did not test them (what changes in the puppy brain during the sensitive period).

3.5 Content Plus Prosody: Benjamin and Slocombe

In Benjamin and Slocombe (2018), two female experimenters sat in the room, each with a loudspeaker that played her own recorded speech. One spoke in naturalistic dog-directed speech, with dog-related content and dog-directed prosody; the other in adult-directed speech with adult content. The measures were looking time during playback and proximity to each person in a one-minute off-lead phase afterwards.

In the first experiment, with 37 dogs from a boarding kennel in York (mean age six years; 33 analyzed for looking, 34 for proximity), the dogs looked significantly longer at the speaker of naturalistic dog-directed speech, with a large effect, and spent more time near her afterwards. In the second experiment, with 32 dogs (29 analyzed for looking), the researchers mismatched content and prosody: dog-related content spoken in adult prosody, and adult content spoken in dog-directed prosody. Now neither speaker was preferred, in looking or in proximity, and a power analysis indicated that the study could have detected an effect.

This result does not show that dogs respond to content only or to tone only. In this set-up, the preference required both together. The authors suggested that dog-directed speech may strengthen the bond between dog and speaker, but they also listed limitations: the same two experimenters' recordings were heard by many dogs, which is pseudoreplication at the level of the stimuli, and in the second experiment the dogs had an a priori preference for one of the experimenters.

3.6 In the Brain: Sensitivity to Exaggerated Prosody

An fMRI study tested 19 trained adult family dogs (two to ten years; golden retrievers, border collies, cocker spaniels, mixed breeds and others) with recordings from 12 female and 12 male speakers the dogs did not know, taken from the Hungarian production study (Gergely et al., 2023). Two non-primary auditory regions in the left hemisphere responded more to naturalistic dog-directed and/or infant-directed speech than to adult-directed speech, especially when the speakers were women. The response increased with mean pitch and pitch variance.

The authors pointed out that the dogs were trained for the scanner, that owner sex and breed were only partly controlled, and that it is unclear whether the effect is a bottom-up response to the acoustic features or a top-down effect of attention. The study shows that dog brains register exaggerated prosody, not that this has consequences for behavior outside the scanner.

4. Words and Intonation: How Dogs Process Speech

4.1 Voices and Emotional Tone

Dogs have their own vocal repertoire and live in a world in which sound carries social information (how dogs communicate with their own voices). A comparative fMRI study reported voice-sensitive regions in the dog brain, with a pattern similar to the anterior temporal voice areas of humans, and sensitivity to the emotional valence of vocal sounds in similarly located non-primary auditory regions (Andics et al., 2014). This paper later received a published erratum (Andics, Gácsi, et al., 2017), which is why the present article does not rely on its original side labels when discussing lateralization.

When dogs heard non-verbal human vocalizations of six basic emotions, head-turn and cardiac measures showed a prevalent right-hemisphere bias for clearly negative ones such as fear and sadness, and a left-hemisphere bias for happiness (Siniscalchi et al., 2018). These were non-verbal vocalizations, not speech, so the study supports only the narrower statement that emotional tone in human voices is processed. Dogs also link voices to individuals: they looked longer when a face did not match the owner's or stranger's voice they had just heard (Adachi et al., 2007). More on how dogs process emotion (what is known about emotions in dogs).

4.2 Andics et al. (2016) and the Published Erratum

The best-known study on words and intonation in dogs used awake fMRI with Hungarian praise words and neutral words, each spoken in a praising and a neutral intonation (Andics et al., 2016). The original abstract reported "a left-hemisphere bias for processing meaningful words, independently of intonation", a right auditory region that distinguished intonationally marked from unmarked words, and stronger activity in reward regions only when both the word and the intonation were consistent with praise.

In 2017, Science published an erratum for this report (Andics, Gábor, et al., 2017). Since then, the same research group has described the preferential processing of praise words as right-lateralized, not left-lateralized. In a follow-up with nearly the same dogs, the authors wrote: "We replicated our earlier findings that in dogs, lexically marked praise words are processed with a right-hemisphere bias" (Gábor et al., 2020). The left-hemisphere claim from the original abstract should therefore not be repeated as a current finding.

The other results in the original abstract, including the reward-region finding, come from one sample of highly trained dogs in one laboratory.

4.3 Gábor et al. (2020): A Right-Hemisphere Bias

The follow-up tested 12 pet dogs (six border collies, five golden retrievers and one German shepherd, aged two to ten years), all but one of whom had taken part in the earlier study (Gábor et al., 2020). The stimuli were three Hungarian praise words ("azaz", "ügyes", "jól van") and three neutral words ("akár", "olyan", "mégsem"), each in praising and neutral prosody.

A classic four-condition analysis found no effects. The adaptation analysis showed short- and long-term adaptation to emotional prosody in subcortical regions, and long-term adaptation to the lexically marked words in cortical auditory regions, with a right-hemisphere bias that varied with age and was independent of prosody in a secondary region. The authors wrote: "we do not propose that dogs have human-analogue lexical representations", and they stated that it is unknown whether the dogs had learned the meaning of the praise words or whether those words had simply become more relevant to them.

4.4 Conflicting Evidence From Head-Turn Studies

In the behavioral head-orienting paradigm, the direction in which a dog turns its head toward a sound is taken as an indication of which hemisphere is processing it more strongly (what laterality in dogs is and how it is measured). Using this method, Ratcliffe and Reby (2014) found a significant left-hemisphere bias when the meaningful phonemic cues in a familiar command were made more salient, and a significant right-hemisphere bias when intonational or speaker-related cues were made more salient.

This is close to the "words left, tone right" pattern known from humans. The corrected fMRI result, with right-biased processing of praise words, does not fit it. Gábor et al. (2020) acknowledged that their result is "harder to reconcile" with such behavioral reports and suggested that orienting biases may not directly reflect hemispheric processing. The prosody study adds a further complication, since the regions that responded to exaggerated prosody were in the left hemisphere (Gergely et al., 2023). The question is open, but a simple "words left, tone right" scheme is not supported by the current fMRI evidence in dogs.

4.5 Dogs Recognize Words Without the Melody

Several studies show that dogs extract information from words, not only from tone. The most direct test used owners' own recordings of a neutral reading text, the so-called rainbow passage, with an embedded phrase that was either meaningful to the dog ("[dog's name], come on then!") or not ("[another name], pass me a coffee!") (Root-Gutteridge et al., 2025). The phrase was spoken either in neutral reading prosody or in dog-directed speech. Owners sat with their back to the dog and wore headphones; the outcome was the dog's gaze at the owner.

For the main study at the University of Sussex, 53 privately owned dogs and a total of 57 participating owners (17 men and 40 women) were recruited; data from 49 dogs remained after exclusions. Dogs looked toward their owners in 99 of 172 analyzed trials (57.6%). Phrase type had a significant effect: dogs were most likely to look when they heard the meaningful phrase in dog-directed speech and least likely when they heard the control phrase in reading prosody. The meaningful phrase in reading prosody did not differ from the control phrase in dog-directed speech, and gaze duration was longest for the meaningful phrase in dog-directed speech. In a second experiment with 13 dogs, owner sex had no significant effect. The authors concluded that dog-directed prosody facilitates the detection of meaningful content by capturing attention, but that dogs also recognize meaningful content without exaggerated prosody.

Other studies point the same way. Dogs preferred their own name over a stress-matched foil spoken by an unfamiliar person, both in quiet and in multi-talker babble down to a signal-to-noise ratio of 0 dB, a level at which, according to the authors, one-year-old human infants fail (Mallikarjun et al., 2019). Working dogs performed better. In a study with vocoded speech, an acoustically degraded form of speech, dogs listened longer to their own name than to foils when it was processed through 16 channels, without prior exposure to this kind of speech (Mallikarjun et al., 2024). In the second experiment, with 25 analyzed dogs, recognition depended on how many phonemes the name contained.

In an older and very small study, six dogs were trained to sit and to come in response to tape-recorded commands, and changes to the phonemes of the commands markedly reduced their performance in all cases (Fukuzawa et al., 2005). The type of change mattered for the altered "sit" command but not for the altered "come" command. Within these limits, the result shows that dogs attend to the sound of the word, not just its melody.

4.6 What Dogs Do Not Hear in Words

There are limits. In an ERP study with 17 awake dogs, brain responses differed between familiar instruction words and phonetically dissimilar nonsense words, but not between known words and phonetically similar nonsense words (Magyari et al., 2020). The difference was larger in dogs who heard the instructions more often. The authors described this as a limited readiness to access phonetic detail, noted that the typical vocabulary of dogs is very low, and called for confirmation in future work.

An fMRI study with 12 dogs trained to retrieve two named objects found stronger activation for novel pseudowords than for the trained words in parietotemporal regions of both hemispheres, which the authors interpreted as novelty detection (Prichard et al., 2018). Only a subset of the dogs had voxel clusters that distinguished the two trained words. Another ERP study found a mismatch response when dogs saw an object that did not match the object word they had just heard, larger for words the owners said their dogs knew well (Boros et al., 2024). This indicates referential expectations about object words, not sentence comprehension. Word learning is covered in a separate article (how dogs learn words).

Dogs can also generalize across speakers. In a habituation and dishabituation study, they spontaneously grouped an unfamiliar speaker across different words, and a word across different speakers of the same gender, which the authors described as an ability that is "not a uniquely human trait" (Root-Gutteridge et al., 2019). In principle, a cue learned from one person can thus be recognized from another (how dogs generalize what they learn). Finally, brain responses in 18 dogs distinguished natural from scrambled speech and produced distinct patterns for a familiar and an unfamiliar language, with larger differences in older dogs (Cuaya et al., 2022). Dogs from English- and Spanish-speaking homes also discriminated the two languages in a head-turn procedure, listening longer to the less familiar one (Mallikarjun et al., 2023). Both show sensitivity to how a language sounds, not comprehension.

5. Tone, Gestures and Context

Tone carries information for dogs even without words. In an object-choice study, an experimenter with her back to the dog produced non-word vocalizations (Colbert-White et al., 2018). In the first experiment, the dogs preferred the container associated with positive, happy-sounding intonation, while negative intonation and a neutral breath did not differ. In the second experiment, with pointing added, the dogs preferred the pointed container and the one associated with negative intonation, and the effect was strongest when both cues were combined. The authors concluded that intonation alone is informative, but that its effect is strongly shaped by pointing (how dogs use human signals to evaluate situations).

In everyday communication, words also compete with the body. Water rescue dogs trained to respond to both gestures and verbal commands followed gestures significantly better when the two were given separately, and when gestures and words conflicted, they mostly followed the gesture (D'Aniello et al., 2016). The exception was the verbal "come" combined with a gestural "stay" while the owner moved away. Females responded more to gestures, while males responded better to verbal commands than females did. How a person moves may count more than what the person says (how dogs read human gestures).

Neither tone nor words work on their own. A cue's effect depends on context, on who is speaking, on what usually follows and on what the dog has experienced with that sound before.

6. Praise and Training

6.1 Is Vocal Praise a Reinforcer?

Whether cheerful vocal praise functions as a reinforcer for an individual dog can depend on its learning history and the situation. The experiments below primarily show that vocal praise was less effective than petting under their test conditions; they do not establish that praise has no reinforcing value for dogs in general. In a series of choice experiments, shelter dogs, owned dogs tested with strangers and owned dogs tested with their owners all preferred petting to vocal praise (Feuerbacher & Wynne, 2015). When only one form of interaction was available, petting produced more proximity than vocal praise, and vocal praise "produced as little proximity-seeking behavior as did no interaction". The dogs did not tire of petting across eight three-minute sessions (what dogs prefer in touch and petting).

An earlier study compared social interaction with food as a reinforcer for a nose-touch task in shelter dogs, owned dogs and hand-reared wolves (Feuerbacher & Wynne, 2012). Brief social interaction, about four seconds of petting with praise, was a relatively ineffective reinforcer compared with food: responding was lower and latencies were longer. The authors concluded that vocal praise "likely has to be specifically conditioned" to work as a reinforcer. The praise studies did not manipulate prosody, so they do not show whether praise spoken in dog-directed speech would have performed differently (how dogs rank food, petting and praise).

6.2 Praise as a Conditioned Signal

Even if vocal praise was only a weak reinforcer on its own in the studies described, it can acquire meaning and reinforcing value through learning. A word that reliably comes before food or play can acquire value as a predictor, just as other conditioned signals do (how reinforcement schedules maintain behavior). This is consistent with the fMRI result that reward regions responded most when both praise words and praising intonation came together (Andics et al., 2016), and with the observation that the brain responses to instruction words were larger in dogs who heard them more often (Magyari et al., 2020). Neither study tested conditioning directly.

In practice: Praise words work best as a marker when they are consistently followed by something the dog values, such as food, play or access to something the dog wants. A praise word that no longer reliably predicts a valued consequence may lose some of its effectiveness. This is an observation from training practice and a consequence of learning theory, not the result of a study on dog-directed speech.

6.3 The Missing Study

No verified study has tested whether dog-directed speech improves learning speed, training success or reliability. The dog studies measured attention, gaze, approach, brain activity or detection, not learning. Borrowing from the infant literature does not help, since the proposed link to language learning runs through vowel hyperarticulation, the very feature that dog-directed speech lacks (Kuhl et al., 1997; Burnham et al., 2002).

7. Popular Claims Checked

7.1 "Dogs Understand Everything We Say"

Not supported. Dogs recognize familiar words, names and phrases, but the evidence points to recognition of relevant sound patterns, not sentence comprehension. The authors of the main fMRI follow-up explicitly do not propose that dogs have human-like lexical representations (Gábor et al., 2020). ERP data show limited attention to phonetic detail (Magyari et al., 2020), and the object-word study shows expectations about particular objects, not understanding of grammar (Boros et al., 2024).

7.2 "Baby Talk Makes Dogs Learn Faster"

Not supported. No verified study has tested it; the dog studies measured attention and related proxies (Ben-Aderet et al., 2017; Jeannin et al., 2017a; Benjamin & Slocombe, 2018; Root-Gutteridge et al., 2025). Attention is a precondition for learning, but more attention in a playback test is not evidence of faster learning.

7.3 "Dogs Respond Only to Tone, Not to Words"

Not supported. Dogs detected a meaningful phrase in a monotonous reading voice (Root-Gutteridge et al., 2025), recognized their names in noise and in vocoded speech (Mallikarjun et al., 2019, 2024), performed worse when phonemes in commands were changed (Fukuzawa et al., 2005) and showed brain responses to known words regardless of intonation (Andics et al., 2016; Prichard et al., 2018; Gábor et al., 2020; Boros et al., 2024). Tone matters, but it is not the whole message.

7.4 "A High Voice Strengthens the Bond"

Only weakly supported. One study using a relevant short-term proximity measure found that kennel dogs spent more time near the speaker of naturalistic dog-directed speech in the minute after playback (Benjamin & Slocombe, 2018). The authors wrote that the register "may" strengthen the bond. There are no long-term data on dog-directed speech and the dog-human relationship.

8. What Follows for Everyday Life

The research does not justify strict rules about how to speak to a dog. There is no need to stop using a high, friendly voice: in some studies it attracted more attention, and it did not interfere with word recognition (Root-Gutteridge et al., 2025). Words matter as well. A dog who has learned a word responds to its sound, and changing the sound can reduce the response (Fukuzawa et al., 2005). And praise is not a reward simply because it sounds happy.

In practice: A raised, friendly voice is useful for getting a dog's attention in everyday life, for example before a cue or when calling the dog from a distance. Once the dog is attending, the cue itself can be spoken in a normal voice. This is an observation from practice, consistent with the finding that dog-directed prosody captures attention, but it has not been tested as a training method.

In practice: Cues are easiest for a dog to recognize when they always sound the same: the same word, a similar tone, and ideally the same words used by everyone in the household. Different people in a family often say "come", "come here" and "here" for the same thing, which turns one cue into three. This is a recommendation from practice, supported indirectly by the finding that phoneme changes disrupted responses to commands.

In practice: Repeating a cue several times in a rising voice can teach the dog that the first repetition may be ignored. It is more useful to say the cue once, wait, and help the dog if needed. This is a recommendation from training practice; no study on dog-directed speech has tested it.

9. Summary at a Glance

Dog-directed speech is a register with a higher mean pitch, usually more pitch variation, a slower word rate and short, repetitive utterances dominated by imperatives and attention-getters. It has been confirmed in several languages (Ben-Aderet et al., 2017; Gergely et al., 2017; Lesch et al., 2019). Its pitch range depends on context and was narrower in a stress test (Lesch et al., 2019; Jeannin et al., 2017b). Unlike infant-directed speech, it lacks vowel hyperarticulation (Burnham et al., 2002; Xu et al., 2013; Gergely et al., 2017).

Results on how dogs respond conflict. One study found that puppies, but not adult dogs, responded more strongly (Ben-Aderet et al., 2017); another found the reverse (Jeannin et al., 2017a). In a live-speaker study, adult dogs preferred naturalistic dog-directed speech, but only when dog-related content and dog-directed prosody came together (Benjamin & Slocombe, 2018). Dogs recognize words without exaggerated prosody, but that prosody makes detection more likely (Root-Gutteridge et al., 2025). The original left-hemisphere claim for word processing was followed by an erratum, and the same group now reports a right-hemisphere bias (Andics, Gábor, et al., 2017; Gábor et al., 2020). Vocal praise alone was a weak reinforcer (Feuerbacher & Wynne, 2012, 2015), and no study has tested whether dog-directed speech improves learning.

10. Research Gaps and Critical Appraisal

10.1 Small Samples

The command study tested 6 dogs (Fukuzawa et al., 2005), the first playback study 10 puppies per experiment (Ben-Aderet et al., 2017), the fMRI studies 12 to 19 dogs and the ERP study 17. On the production side, the first description of the register was based on four owner-dog pairs (Hirsh-Pasek & Treiman, 1982), and key acoustic comparisons used 9 to 12 speakers (Burnham et al., 2002; Jeannin et al., 2017a; Xu et al., 2013). Such samples are enough to detect large effects, but not to estimate how common an effect is or how much individual dogs differ.

10.2 Proxy Measures

The outcome measures are gaze duration, head-orienting, composite behavioral scores, proximity during one minute and BOLD signal. None measures learning, obedience, stress or the long-term relationship between dog and person. The gap between "the dog looked longer" and "the dog benefits" is not bridged by any study reviewed here.

10.3 Ecological Validity

Speech was recorded in front of photographs (Ben-Aderet et al., 2017), as scripted sentences, or as a read text with an embedded phrase. Playback mostly happened through loudspeakers without a visible person. The one study with live speakers had to accept pseudoreplication, because two experimenters' recordings were heard by many dogs (Benjamin & Slocombe, 2018). Everyday conversation between dog and owner has hardly been studied.

10.4 Language and Culture

The production and perception studies come from English-speaking (United States, United Kingdom, Australia), French, Hungarian and Austrian samples, with one study from English- and Spanish-speaking homes. No verified production study comes from a non-Western culture. Whether the register is general or culture-specific is open, as Benjamin and Slocombe (2018) noted. The conflicting findings on pitch in dog-directed versus infant-directed speech between Australian and Hungarian speakers point in the same direction (Burnham et al., 2002; Gergely et al., 2017).

10.5 Who Speaks and Which Dogs Listen

Female voices dominate the stimulus sets: 30 women in one study (Ben-Aderet et al., 2017), 9 in another (Jeannin et al., 2017a), two female experimenters in a third (Benjamin & Slocombe, 2018). One fMRI study found stronger responses to female speakers (Gergely et al., 2023), whereas owner sex had no significant effect in the reading study (Root-Gutteridge et al., 2025). The fMRI dogs are highly trained and drawn from few breeds, mainly border collies and golden retrievers. Results from these and from shelter and kennel populations may not transfer directly to family dogs at home.

10.6 Corrections and Replication

Two fMRI studies on dogs and voices received published errata (Andics, Gábor, et al., 2017; Andics, Gácsi, et al., 2017). The 2020 follow-up describes its right-hemisphere finding as a replication, but it was carried out by the same group with almost the same dogs (Gábor et al., 2020). Independent replication, ideally in other laboratories, other languages and less selected dogs, is still missing for the fMRI findings, and the conflict between fMRI and head-turn results remains unresolved.

11. Conclusion

People talk to dogs in a recognizable register: higher pitch, more melody, slower and more repetitive speech. The most telling detail is what is missing. People exaggerate their vowels for infants, who are expected to learn language, but not for dogs. Dog-directed speech looks less like teaching and more like a way of reaching a listener who responds socially but does not talk back.

Whether dogs profit from it is less clear. The results for puppies and adult dogs contradict each other. Well supported is that dogs do not depend on the melody: they recognize words in a flat voice, in noise and in degraded speech, although the high voice can help them notice. Praise in a happy voice is not automatically a reward, and no study has shown that baby talk makes dogs learn faster. The practical lesson is modest: talk to your dog in whatever voice feels natural, keep cues consistent, and remember that what a sound means to a dog depends on what usually follows it.

Key Insights (Takeaways)

• Dog-directed speech has a higher pitch, usually more pitch variation, a slower word rate and many short, repetitive imperatives; this has been shown in several languages (Ben-Aderet et al., 2017; Gergely et al., 2017; Benjamin & Slocombe, 2018).

• Unlike speech to infants, speech to dogs does not exaggerate vowels, which the authors interpret as a sign that people do not treat dogs as language learners (Burnham et al., 2002; Xu et al., 2013).

• Findings on age conflict: one small study found stronger responses in puppies, another found them in adult dogs (Ben-Aderet et al., 2017; Jeannin et al., 2017a).

• In a live-speaker study, adult dogs preferred dog-directed speech only when dog-related content and dog-directed prosody came together (Benjamin & Slocombe, 2018).

• Dogs recognize meaningful words without exaggerated prosody, but dog-directed prosody makes detection more likely (Root-Gutteridge et al., 2025).

• An erratum was published for the 2016 word-processing study; the same group now reports a right-hemisphere bias for praise words and calls it a replication (Andics, Gábor, et al., 2017; Gábor et al., 2020).

• In the tested procedures, vocal praise alone was a weaker reinforcer than petting or food, and its value may depend on learning history and context (Feuerbacher & Wynne, 2012, 2015). This does not demonstrate that dog-directed speech improves learning.

References

Adachi, I., Kuwahata, H., & Fujita, K. (2007). Dogs recall their owner's face upon hearing the owner's voice. Animal Cognition, 10(1), 17–21. https://doi.org/10.1007/s10071-006-0025-8

Andics, A., Gábor, A., Gácsi, M., Faragó, T., Szabó, D., & Miklósi, Á. (2016). Neural mechanisms for lexical processing in dogs. Science, 353(6303), 1030–1032. https://doi.org/10.1126/science.aaf3777

Andics, A., Gábor, A., Gácsi, M., Faragó, T., Szabó, D., & Miklósi, Á. (2017). Erratum for the Report "Neural mechanisms for lexical processing in dogs". Science, 356(6333), eaan3276. https://doi.org/10.1126/science.aan3276

Andics, A., Gácsi, M., Faragó, T., Kis, A., & Miklósi, Á. (2014). Voice-sensitive regions in the dog and human brain are revealed by comparative fMRI. Current Biology, 24(5), 574–578. https://doi.org/10.1016/j.cub.2014.01.058

Andics, A., Gácsi, M., Faragó, T., Kis, A., & Miklósi, Á. (2017). Voice-sensitive regions in the dog and human brain are revealed by comparative fMRI (erratum). Current Biology, 27(8), 1248–1249. https://doi.org/10.1016/j.cub.2017.03.036

Ben-Aderet, T., Gallego-Abenza, M., Reby, D., & Mathevon, N. (2017). Dog-directed speech: Why do we use it and do dogs pay attention to it? Proceedings of the Royal Society B: Biological Sciences, 284(1846), 20162429. https://doi.org/10.1098/rspb.2016.2429

Benjamin, A., & Slocombe, K. (2018). 'Who's a good boy?!' Dogs prefer naturalistic dog-directed speech. Animal Cognition, 21(3), 353–364. https://doi.org/10.1007/s10071-018-1172-4

Boros, M., Magyari, L., Morvai, B., Hernández-Pérez, R., Dror, S., & Andics, A. (2024). Neural evidence for referential understanding of object words in dogs. Current Biology, 34(8), 1750–1754.e4. https://doi.org/10.1016/j.cub.2024.02.029

Burnham, D., Kitamura, C., & Vollmer-Conna, U. (2002). What's new, pussycat? On talking to babies and animals. Science, 296(5572), 1435. https://doi.org/10.1126/science.1069587

Colbert-White, E. N., Tullis, A., Andresen, D. R., Parker, K. M., & Patterson, K. E. (2018). Can dogs use vocal intonation as a social referencing cue in an object choice task? Animal Cognition, 21(2), 253–265. https://doi.org/10.1007/s10071-018-1163-5

Cuaya, L. V., Hernández-Pérez, R., Boros, M., Deme, A., & Andics, A. (2022). Speech naturalness detection and language representation in the dog brain. NeuroImage, 248, 118811. https://doi.org/10.1016/j.neuroimage.2021.118811

D'Aniello, B., Scandurra, A., Alterisio, A., Valsecchi, P., & Prato-Previde, E. (2016). The importance of gestural communication: A study of human–dog communication using incongruent information. Animal Cognition, 19(6), 1231–1235. https://doi.org/10.1007/s10071-016-1010-5

Feuerbacher, E. N., & Wynne, C. D. L. (2012). Relative efficacy of human social interaction and food as reinforcers for domestic dogs and hand-reared wolves. Journal of the Experimental Analysis of Behavior, 98(1), 105–129. https://doi.org/10.1901/jeab.2012.98-105

Feuerbacher, E. N., & Wynne, C. D. L. (2015). Shut up and pet me! Domestic dogs (Canis lupus familiaris) prefer petting to vocal praise in concurrent and single-alternative choice procedures. Behavioural Processes, 110, 47–59. https://doi.org/10.1016/j.beproc.2014.08.019

Fukuzawa, M., Mills, D. S., & Cooper, J. J. (2005). The effect of human command phonetic characteristics on auditory cognition in dogs (Canis familiaris). Journal of Comparative Psychology, 119(1), 117–120. https://doi.org/10.1037/0735-7036.119.1.117

Gábor, A., Gácsi, M., Szabó, D., Miklósi, Á., Kubinyi, E., & Andics, A. (2020). Multilevel fMRI adaptation for spoken word processing in the awake dog brain. Scientific Reports, 10, 11968. https://doi.org/10.1038/s41598-020-68821-6

Gergely, A., Faragó, T., Galambos, Á., & Topál, J. (2017). Differential effects of speech situations on mothers' and fathers' infant-directed and dog-directed speech: An acoustic analysis. Scientific Reports, 7, 13739. https://doi.org/10.1038/s41598-017-13883-2

Gergely, A., Gábor, A., Gácsi, M., Kis, A., Czeibert, K., Topál, J., & Andics, A. (2023). Dog brains are sensitive to infant- and dog-directed prosody. Communications Biology, 6, 859. https://doi.org/10.1038/s42003-023-05217-y

Hirsh-Pasek, K., & Treiman, R. (1982). Doggerel: Motherese in a new context. Journal of Child Language, 9(1), 229–237. https://doi.org/10.1017/S0305000900003731

Jeannin, S., Gilbert, C., Amy, M., & Leboucher, G. (2017a). Pet-directed speech draws adult dogs' attention more efficiently than adult-directed speech. Scientific Reports, 7, 4980. https://doi.org/10.1038/s41598-017-04671-z

Jeannin, S., Gilbert, C., & Leboucher, G. (2017b). Effect of interaction type on the characteristics of pet-directed speech in female dog owners. Animal Cognition, 20(3), 499–509. https://doi.org/10.1007/s10071-017-1077-7

Kuhl, P. K., Andruski, J. E., Chistovich, I. A., Chistovich, L. A., Kozhevnikova, E. V., Ryskina, V. L., Stolyarova, E. I., Sundberg, U., & Lacerda, F. (1997). Cross-language analysis of phonetic units in language addressed to infants. Science, 277(5326), 684–686. https://doi.org/10.1126/science.277.5326.684

Lesch, R., Kotrschal, K., Schöberl, I., Beetz, A., Solomon, J., & Fitch, W. T. (2019). Talking to dogs: Companion animal-directed speech in a stress test. Animals, 9(7), 417. https://doi.org/10.3390/ani9070417

Magyari, L., Huszár, Z., Turzó, A., & Andics, A. (2020). Event-related potentials reveal limited readiness to access phonetic details during word processing in dogs. Royal Society Open Science, 7(12), 200851. https://doi.org/10.1098/rsos.200851

Mallikarjun, A., Shroads, E., & Newman, R. S. (2019). The cocktail party effect in the domestic dog (Canis familiaris). Animal Cognition, 22(3), 423–432. https://doi.org/10.1007/s10071-019-01255-4

Mallikarjun, A., Shroads, E., & Newman, R. S. (2023). Language preference in the domestic dog (Canis familiaris). Animal Cognition, 26(2), 451–463. https://doi.org/10.1007/s10071-022-01683-9

Mallikarjun, A., Shroads, E., & Newman, R. S. (2024). Perception of vocoded speech in domestic dogs. Animal Cognition, 27(1), 34. https://doi.org/10.1007/s10071-024-01869-3

Mitchell, R. W. (2001). Americans' talk to dogs: Similarities and differences with talk to infants. Research on Language and Social Interaction, 34(2), 183–210. https://doi.org/10.1207/S15327973RLSI34-2_2

Mitchell, R. W. (2004). Controlling the dog, pretending to have a conversation, or just being friendly? Influences of sex and familiarity on Americans' talk to dogs during play. Interaction Studies, 5(1), 99–129. https://doi.org/10.1075/is.5.1.06mit

Mitchell, R. W., & Edmonson, E. (1999). Functions of repetitive talk to dogs during play: Control, conversation, or planning? Society & Animals, 7(1), 55–81. https://doi.org/10.1163/156853099X00167

Prato-Previde, E., Fallani, G., & Valsecchi, P. (2006). Gender differences in owners interacting with pet dogs: An observational study. Ethology, 112(1), 64–73. https://doi.org/10.1111/j.1439-0310.2006.01123.x

Prichard, A., Cook, P. F., Spivak, M., Chhibber, R., & Berns, G. S. (2018). Awake fMRI reveals brain regions for novel word detection in dogs. Frontiers in Neuroscience, 12, 737. https://doi.org/10.3389/fnins.2018.00737

Ratcliffe, V. F., & Reby, D. (2014). Orienting asymmetries in dogs' responses to different communicatory components of human speech. Current Biology, 24(24), 2908–2912. https://doi.org/10.1016/j.cub.2014.10.030

Root-Gutteridge, H., Korzeniowska, A., Ratcliffe, V., & Reby, D. (2025). Domestic dogs (Canis familiaris) recognise meaningful content in monotonous streams of read speech. Animal Cognition, 28(1), 29. https://doi.org/10.1007/s10071-025-01948-z

Root-Gutteridge, H., Ratcliffe, V. F., Korzeniowska, A. T., & Reby, D. (2019). Dogs perceive and spontaneously normalize formant-related speaker and vowel differences in human speech sounds. Biology Letters, 15(12), 20190555. https://doi.org/10.1098/rsbl.2019.0555

Siniscalchi, M., d'Ingeo, S., Fornelli, S., & Quaranta, A. (2018). Lateralized behavior and cardiac activity of dogs in response to human emotional vocalizations. Scientific Reports, 8, 77. https://doi.org/10.1038/s41598-017-18417-4

Xu, N., Burnham, D., Kitamura, C., & Vollmer-Conna, U. (2013). Vowel hyperarticulation in parrot-, dog- and infant-directed speech. Anthrozoös, 26(3), 373–380. https://doi.org/10.2752/175303713X13697429463592