"Tat" for cat. "Tiss" for kiss. "Doe" for go, "dum" for gum, "o-day" for okay.
And then the sentence parents almost always add: but he can say it sometimes. He says "book." He says "dog." He says "sticker" perfectly. So it cannot be that he is unable to make the sound — can it?
Both halves of that are worth taking seriously, and they point the same direction. /k/ and /g/ do not behave like the sounds families usually ask us about. They are not late sounds. They are early ones, and that changes the answer to almost every question that follows.
/k/ and /g/ Are Early Sounds, Not Late Ones
The best available figures for American English come from Crowe and McLeod's 2020 review of children's English consonant acquisition in the United States (American Journal of Speech-Language Pathology), which pooled 15 studies covering 18,907 children. Using a 90% criterion — the age by which nine in ten children produce the sound correctly:
- /b, n, m, p, h, w, d/ — acquired by 2;0 to 2;11 (years;months)
- /g, k, f, t, ng, y/ — acquired by 3;0 to 3;11
- /v, j, s, ch, l, sh, z/ — acquired by 4;0 to 4;11
- /r/, voiced "th," "zh" — acquired by 5;0 to 5;11
- voiceless "th" — acquired by 6;0 to 6;11
The authors' own summary is that "on average, all plosives, nasals, and glides were acquired by 3;11." /k/ and /g/ are plosives. They land in the second band, a full two years ahead of /r/ and three ahead of the "th" of thumb.
ASHA arrives at a similar place from a different direction. Its selected phonological patterns resource gives a likely age of elimination of 4 for fronting, the pattern behind these errors.
That matters because of what it rules out. For /r/ and the TH sounds, "let us give it another year" is frequently the correct clinical answer, and we say so on those pages. Here it usually is not. A four-year-old still saying "tat" for cat is past the published window, not approaching it.
One sentence is worth weighing carefully. In their 2021 study of this exact pattern, Cleland and Scobbie note that "absence of velars in the phonetic inventory at 3 years of age is predictive of phonological disorder," citing Stoel-Gammon (1996). A three-year-old with no /k/ or /g/ anywhere is not simply a late bloomer on one sound.
None of this means a single number decides anything. ASHA's own consonant age of acquisition handout attaches Storkel's (2019) caution to the data: "Developmental norms are one source of information," and "an accurate diagnosis can only be achieved by combining multiple measures to converge on a diagnosis." The age tells you whether to get it looked at, not what the answer will be.
"But He Can Say K Sometimes"
This is the observation parents trust least and should trust most. It is real, it is well described, and it has a name.
In the pattern called positional velar fronting, fronting "applies only in the onset of an initial or stressed syllable." McAllister Byun's 2012 study in the Journal of Child Language, Positional velar fronting: an updated articulatory account, followed one child from 3;9 to 4;4 and recorded all four halves of the pattern in a single session at 4;3:
- Word-initial velars fronted — "can" produced as [den], "going" as [doi], "ghost" as [dos], "card" as [dout].
- Word-final velars not fronted — "look," "cake," "like" and "dog" all keeping their final K or G.
- Word-medial velars before a stressed vowel fronted — "okay" as [o-DAY], "because" as [bi-DUS], "again" as [a-DEN], "raccoon" as [wa-TUN].
- Word-medial velars before an unstressed vowel not fronted — "sticker," "bucket," "doggie" and "licking" all intact.
Look at "cake." The K at the end survived; the K at the start did not. Same sound, same word, same child, same breath.
Byun's explanation is mechanical rather than mysterious: children "have difficulty executing discrete tongue movements, preferring to move the tongue and jaw as a single unit," and that difficulty costs more in the motorically demanding position at the front of a word or stressed syllable than it does in a weak one. The tongue is being asked to do something harder in the places where it fails.
Two practical consequences. First, a child who does this is not being lazy or careless, and asking them to "try harder on cat" is asking for the one thing the pattern predicts they cannot yet do on demand. Second — and this is genuinely good news — if the gesture shows up anywhere, it exists. Therapy is relocating something, not building it from nothing.
The Error Patterns Behind "Tat" and "Doe"
- Fronting. ASHA defines it as "a sound made in the back of the mouth (velar) is replaced with a sound made in the front of the mouth (e.g., alveolar)," and its own two examples are "/tæt/ for cat" and "/deɪt/ for gate." Likely age of elimination: 4. This is by far the most common thing parents are hearing.
- Deletion. The sound is left out rather than replaced — "at" for cat, "o" for go, or a book that ends after "boo." Final consonant deletion is its own pattern in ASHA's list and often travels with other simplifications.
- One sound and not the other. A child may front /k/ while keeping /g/, or the reverse. The two differ only in voicing, so a split usually tells you something specific about where the breakdown sits.
- Backing. The reverse direction: tea arriving as "key," do as "goo," two as "coo." One honest footnote here — ASHA's published list includes fronting but not its reverse, and the page states that the list "is not exhaustive," so you may not find backing named on a report or in a parent handout. Clinically it is generally treated as the less expected direction and as a reason to have speech looked at rather than monitored. We are not aware of a solid published prevalence figure for it and are not going to invent one.
- Something in between. Not quite a /t/, not quite a /k/, and transcribed differently by different listeners. That one has its own section below, because it is the version most likely to stall.
Where a whole class of back sounds is being moved forward by a rule, the diagnosis is usually a phonological disorder rather than an articulation disorder, and the treatment differs accordingly. Our comparison of phonological vs. articulation disorder walks through how an evaluation tells them apart.
What It Costs in Intelligibility
More than parents expect, because /k/ and /g/ are frequent and because fronting them in initial position does not produce nonsense — it produces other real words.
Key becomes tea. Cap becomes tap. Cot becomes tot, coat becomes tote, cold becomes told, came becomes tame, call becomes tall. Gate becomes date, got becomes dot, gum becomes dumb, game becomes dame. Cleland and Scobbie put it plainly: velar fronting leads to "homophony and breakdowns in communication." A listener who mishears "tat" as an attempt at cat has done well; a listener who hears a perfectly good "tap" has no signal that anything went wrong at all.
For a sense of scale, ASHA's articulation and phonology portal collects two sets of intelligibility benchmarks. For unfamiliar listeners hearing context-free single words from monolingual American English children (Hustad et al., 2021): 50% understood at 31 months, 75% at 49 months, 90% at 83 months. For parent estimates of children without speech sound disorders (Coplan & Gleason, 1988): 50% at 22 months, 75% at 37 months, 100% at 47 months. A four-year-old who is still noticeably hard for strangers to follow is worth measuring against those numbers rather than against a sibling.
Why Some Children Get Stuck
For most children, this resolves. Cleland and Scobbie describe children's velar fronting as "normally transient," and note that it is "normally successfully treated with phonological interventions." The interesting question is the minority for whom it is not, and there is now direct articulatory evidence about what is happening in those children.
Their 2021 paper in the Journal of Speech, Language, and Hearing Research, The Dorsal Differentiation of Velar From Alveolar Stops in Typically Developing Children and Children With Persistent Velar Fronting, used ultrasound tongue imaging on 30 typically developing children to establish norms for how far apart /k/ and /t/ actually sit in the mouth, then applied those measures to five children with persistent velar fronting who were receiving ultrasound visual biofeedback.
Before intervention, those five children showed values "near zero," which the authors describe as "a complete merger between /k/ and /t/." During intervention the values became variable. After intervention they fell "within the range of typical children." The conclusion the authors draw is that "/k/ is acquired in an articulatorily gradient manner" — not flipped on like a switch, but inched into place.
Two ideas from that literature are worth knowing as a parent. One is undifferentiated lingual gestures, described by Gibbon (1999) in 12 children with functional articulation disorders: a single broad tongue-to-palate contact that is not really a /t/ or a /k/, but which a listener will transcribe as one or the other. The other is covert contrast — "despite a perceptual neutralization, there is any measurable acoustic or articulatory difference between attempts at different phonemes." A child can be making a distinction that no adult ear can hear, which means they have the phonological knowledge and are failing only at the execution.
The detail that should carry the most weight for a parent deciding whether to wait: those five children with persistent velar fronting were between 6;1 and 13;2. Not four. Not five.
Is Tongue Thrust Behind It? Usually Not
Most pages on this site end up somewhere near orofacial myofunctional therapy, because for a great many sound errors the resting position of the tongue is genuinely part of the picture. This is the clearest exception, and it would be dishonest to blur it.
ASHA's orofacial myofunctional disorders portal lists the speech signs associated with these disorders as "distorted productions of /s, z/, often with an interdental lisp, or abnormal lingual dental articulatory placement for /t, d, l, n, ʧ, ʤ, ʃ, ʓ/." /k/ and /g/ do not appear on that list — and they could not sensibly appear on it, because every sound named there is made at or near the teeth and the alveolar ridge, while /k/ and /g/ are made at the soft palate, as far back as English consonants go. A tongue that habitually rests low and forward is sitting in the wrong place for /s/ and for the "sh" family. It is not sitting in the wrong place for /k/.
What matters for the velars is a different capacity: whether the back of the tongue can lift while the front stays down, independently of the jaw. That is a differentiation problem, and it is exactly what both Byun's "move as a unit" account and Cleland and Scobbie's ultrasound data are describing.
A child can of course have both — a tongue thrust and a velar error — and both deserve attention. They are simply not the same finding, and ASHA is explicit that clinicians should "exercise caution before using OMT as a standalone approach to address speech sound errors." If you have been told that tongue thrust is the reason for "tat," it is fair to ask how. Our comparison of myofunctional therapy vs. oral motor therapy covers where the two kinds of work genuinely do overlap.
What Therapy Involves
Because fronting is a rule applied across a class of sounds rather than one sound produced incorrectly, the usual starting point is a phonological approach rather than sound-by-sound drill. ASHA describes two of the common ones:
- Minimal oppositions, which "uses pairs of words that differ by only one phoneme or single feature signaling a change in meaning." For this pattern the pairs are already sitting there: key/tea, cap/tap, gate/date. The point is not repetition — it is showing the child that the difference carries meaning, which is precisely the information a merged contrast is missing.
- The cycles approach, which "targets phonological pattern errors" and "is designed for children with highly unintelligible speech." Where several patterns are stacked and a child is hard to follow, cycling through them tends to buy intelligibility faster than perfecting one sound first.
Where to start is usually decided by stimulability — in ASHA's words, "the child's ability to accurately imitate a model of the target speech sound." This is where the positional pattern earns its keep. A child who already produces final /k/ in book has the dorsal gesture; the work is moving it to the front of a word, which is a different and generally faster job than establishing it from scratch. A child with no velars anywhere starts further back, with the back of the tongue learning to lift on its own.
For the small group that does not respond to phonological intervention, the Cleland and Scobbie work points at why: if the contrast is articulatorily merged rather than phonologically absent, the child needs to see or feel the movement, not hear more models of it. That is the rationale behind biofeedback approaches, and it is a reasonable question to raise if months of good therapy have not moved the sound.
Getting an Evaluation
An evaluation for /k/ and /g/ answers a short list of questions: which of the two sounds is affected, in which word positions, whether the error is a substitution or a deletion, whether the pattern is positional, whether the child is stimulable for the sound anywhere, and whether other phonological patterns are stacked on top. It also checks what is traveling alongside — ASHA reports, citing Eadie et al. (2015), that "40% of children with SSDs had concomitant language impairment," which changes the plan.
Because these are early sounds, the timing conversation here is shorter than it is for the later ones. Our guide to the signs your child may need speech therapy covers what else an evaluation weighs, our speech milestones by age guide puts these sounds in sequence, and our articulation disorder treatment page describes the full range of sound errors we work on as part of speech therapy at Bloom. If the sound in question is /l/ rather than /k/, our guide to L sound speech therapy is the nearest companion; for children generally, see speech therapy for children.
Sessions are available in person at our Dallas office and through online speech therapy nationwide.
Frequently Asked Questions
At what age should a child be able to say K and G?
Earlier than most of the sounds parents ask about. Crowe and McLeod's 2020 review of 15 studies covering 18,907 children in the United States places /g/ and /k/ in the 3;0-3;11 band under a 90% criterion, and summarizes that on average all plosives, nasals and glides were acquired by 3;11. ASHA separately gives 4 as the likely age of elimination for fronting. A four-year-old still saying "tat" for cat has passed the window — a very different situation from /r/ (5;0-5;11) or the voiceless "th" (6;0-6;11).
Why can my child say "book" and "dog" but not "cat"?
Because velar fronting is often positional. In the pattern described as positional velar fronting, fronting applies only in the onset of an initial or stressed syllable. McAllister Byun's 2012 case study recorded exactly that in one four-year-old: "can" as [den] and "going" as [doi], while "look" and "dog" were correct; "okay" became [o-day] while "sticker" and "bucket" stayed intact. The same child produced "cake" as [dak] — the final K survived, the initial one did not. It is a predictable pattern rather than inconsistency, and the fact that the gesture exists somewhere in the word is useful for therapy.
Will my child outgrow velar fronting?
Many children do, and quickly — Cleland and Scobbie call children's velar fronting "normally transient." But age matters. Their paper notes that absence of velars in the phonetic inventory at 3 years of age is predictive of phonological disorder, citing Stoel-Gammon (1996), and the five children with persistent velar fronting in their own study were between 6;1 and 13;2. "Wait and see" is often right for /r/ or "th." It is a much weaker answer here.
Is tongue thrust causing my child's K and G errors?
Usually not. ASHA's list of speech signs under orofacial myofunctional disorders names distorted /s/ and /z/ and abnormal lingual dental placement for /t, d, l, n, ch, j, sh, zh/ — /k/ and /g/ are not on it, and would not fit, because those are front-of-the-mouth placement problems while /k/ and /g/ are made at the soft palate. What matters for velars is whether the back of the tongue can lift independently of the front and of the jaw. A child can have both a tongue thrust and a velar error, but one does not explain the other.
What does therapy for K and G actually look like?
Usually a phonological approach rather than sound-by-sound drill, because fronting is a rule affecting a class of sounds. Cleland and Scobbie note that velar fronting is normally successfully treated with phonological interventions. ASHA describes minimal oppositions, which "uses pairs of words that differ by only one phoneme or single feature signaling a change in meaning" (key/tea does that work directly), and the cycles approach, which "targets phonological pattern errors" and "is designed for children with highly unintelligible speech." Stimulability testing usually decides where to start.
Can K and G sound therapy be done online?
Yes. /k/ and /g/ have no visible contact point — nothing useful happens at the lips or teeth — so very little is lost working through a screen. The sound is taught by listening, by contrast, and by shaping from what the child can already do. Bloom provides online speech therapy nationwide alongside in-person sessions in Dallas.
Hearing "Tat" for "Cat"?
Unlike /r/ and "th," these sounds should be in place well before kindergarten — so the useful question is not whether to wait but what exactly is going on. Schedule a free consultation with Laura Friedman, MS, CCC-SLP, QOM — in person in Dallas or online nationwide — to find out which sounds and positions are affected, whether the pattern is positional, and whether the contrast is simply late or genuinely merged.