Reading Errors Are Only Half the Story of Reading Aloud
12 Oct 2026

Reading Errors Are Only Half the Story

Adrian DeWitts
Written By Adrian DeWitts

Listen to a child read aloud and the page tells only half the story. Across 4,446 children reading Bookbot’s English books, we counted about 36 departures from the text for every 100 words. Half were reading errors that left a word wrong. The other half (repeats, re-reads, fillers, half-started words) left the word right, and a standard fluency test does not score them as errors.

Key Findings

  • Stumbles are as common as mistakes. An error leaves a word wrong (“shop” for “ship”). A disfluency is a stumble on the way to the right word: repeating it, re-reading, an “um” or a half-started “b- ball”. Children made about as many of each, 17 to 18 per 100 words.
  • Children stumble where words get hard. Part-word starts (“b- ball”) were five times as common on long words as on short ones.
  • Pauses flag trouble. A word read after a half-second pause was misread 21% of the time, against 14% after a shorter gap.
  • Sound slips gather at word ends and in vowels.
  • It can be measured at scale. A speech recognition system that keeps partial words agreed with a human listener on about 9 in 10 of the commonest events.

The recordings lean toward harder reading, so these rates describe; they are not norms.

Why We Counted

Reading miscue analysis is usually done by hand, and hand-coded studies are small. Nguyen, Del Tufo and Cutting (2020) coded the miscues of 82 children; Miller and Schwanenflugel (2008) followed the pauses of 92. Standard scoring narrows the view further: DIBELS ignores repetitions and inserted words and scores a quick self-correction as correct (University of Oregon, 2026), so most reading disfluency is never scored as an error.

How We Measured Reading Errors

Standard speech recognition tidies speech into clean words and drops partial words and fillers. We developed a speech recognition system for children’s oral reading that also keeps partial words (“c-”), repetitions and fillers, aiming for a verbatim transcript.

Each transcript was then aligned word by word against the book’s text. The alignment allows for partial pages, re-reading, skipping, contractions and split words. Homophones count as correct, since the spelling is the transcriber’s. Each book word’s last reading is its outcome; an earlier wrong reading of the same word is a self-correction.

Every remaining difference was classified. Substitutions are sorted by how alike the two words are in spelling and sound, using a children’s pronunciation dictionary. Repetitions and self-corrections are identified from what the child said around them, and each partial word is tied to the word it starts. Pause lengths come from forced alignment, which times each word in the audio.

Rates are per 100 book words attempted, from the first to the last book word a recording reached, omissions included. Words cut off at a recording’s end are not counted against the child.

We checked 147 recordings by ear; for the commonest kinds, about 9 in 10 transcribed events were confirmed. The recordings analyzed are those selected for full verbatim transcription because they were hardest to transcribe. They lean toward harder reading: children read 82% of book words correctly in them, against 94% in the remaining recordings.

Four-step method diagram: a child reads aloud, the speech recognition system writes down every word, part-word and filler, the transcript is aligned to the book text, and each difference is sorted into a kind such as repeat, part-word start or near miss

Types of Reading Errors and Disfluencies, Kind by Kind

Reading errors and disfluencies, by kind per 100 book words

Pick a kind in the menu below the bars, or click a bar, to see every form counted for it.

The word ends up wrong

Near miss ship → shop7.54
Different word cat → dog4.35
Small-word swap the → a3.22
Ending changed jumped → jump1.36
Small word left out0.48
Content word left out0.19
Voicing swap pig → big0.18
Reversal saw → was0.09
3+ words skipped0.05

Disfluencies

Word repeated the the6.47
Extra word or remark2.36
Phrase re-read the cat the cat1.78
Extra small word1.67
Filler um, uh, er1.47
Word from elsewhere on the page1.29
Part-word start b- ball1.21
Self-correction sit → sat0.77
Stray part-word sound0.57
Fixed on re-reading0.20
Sounding out c- a- t0.16
Started, then another word0.09
Start after the word dog do-0.07

Pauses

Pause 0.5–1 s14.06
Pause 1 s or more13.92

Counted as correct

Homophone or spacing to → two1.12
Contraction do not → don't0.16

Near miss ship → shop

A word close to the book word in spelling or sound, or a non-word.

#ExampleChildrenTimes
1is → it's136172
2it → it's120157
3pup → pop6287
4it → eat6274
5can → can't5763
6cod → cold5269
7she → see5065
8will → well4955
9they → day3943
10his → he's3845
11spots → sports3741
12wants → one3637
13can → cat3439
14ran → run3343
15he → hey3338
16cod → code3336
17says → said3240
18spot → sport3136
19dad → that3035
20her → hair3035

Different word cat → dog

#ExampleChildrenTimes
1a → all3342
2hat → head1922
3a → yeah1826
4to → go1818
5all → a1717
6a → say1616
7zom → some1522
8the → that's1515
9a → cat1313
10does → don't1313
11fish → face1214
12got → good1213
13help → have1213
14a → day1212
15little → the1112
16a → air1013
17a → eat1011
18is → see1010
19is → eats1010
20is → he's1010

Small-word swap the → a

#ExampleChildrenTimes
1a → the267394
2the → a178233
3the → that166214
4in → and135174
5a → and132149
6not → no98110
7is → I95104
8she → he87105
9he → she79106
10off → of7989
11it → I7981
12the → they7899
13of → off7892
14it → is7388
15the → this7387
16can → and7183
17they → the7077
18a → at6577
19is → it6570
20this → the6270

Ending changed jumped → jump

#ExampleChildrenTimes
1gets → get81104
2has → had5156
3wants → want4451
4lot → lots4252
5says → say4150
6puts → put3741
7get → gets3540
8likes → like3437
9that → that's3337
10birds → bird3334
11cat → cats3033
12spots → spot3030
13starts → start2932
14finds → find2930
15sings → sing2832
16had → has2730
17makes → make2629
18bird → birds2124
19stands → stand2122
20lights → light2030

Small word left out

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book wordChildrenTimes
1a176226
2the120183
3it7597
4and6892
5to5966
6is5661
7in5563
8that4451
9of3846
10up3339
11I3236
12he3134
13she2425
14are2227
15at2123
16not2122
17with2023
18so1921
19on1920
20will1720

Content word left out

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book wordChildrenTimes
1all2325
2still1012
3like1010
4just910
5big78
6hmm78
7says78
8fish77
9fun66
10one66
11back66
12bee56
13get56
14little55
15pee45
16gets45
17time45
18ah44
19oh44
20lot44

Voicing swap pig → big

#ExampleChildrenTimes
1pig → big3844
2ned → net2738
3rot → rod1623
4lamp → lamb1518
5cod → cot1416
6thinks → things1112
7put → but1111
8got → god1012
9hat → had1010
10live → life911
11sings → sinks910
12dibs → dips812
13bop → bob810
14rod → rot89
15pat → bat710
16life → live79
17sat → sad78
18big → pig77
19things → thinks77
20zing → sing68

Reversal saw → was

#ExampleChildrenTimes
1dig → big2931
2spot → stop1519
3lost → lots1011
4Dan → and99
5being → begin89
6from → form77
7quite → quiet67
8snug → sung66
9spots → stops66
10bad → dad56
11lots → lost55

3+ words skipped

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book text skippedChildrenTimes
1j o j22
2a beak a11
3a bit of11
4no it is11
5down on a11
6a stag the11
7a toy bike11
8he picks a11
9he took it11
10she is mad11
11so good at11
12and held on11
13can pant if11
14first it is11
15he wants to11
16I dont know11
17I would not11
18likely as a11
19off the bus11
20they go for11

Word repeated the the

#ExampleChildrenTimes
1the the333628
2it it280579
3a a258444
4I I183331
5and and179303
6no no175375
7is is155232
8he he154230
9to to147225
10in in114188
11up up101257
12she she95171
13that that95152
14what what94170
15they they94125
16but but91167
17can can90138
18not not90136
19this this86151
20big big84135

Extra word or remark

#ExampleChildrenTimes
1it's104119
2one7176
3yeah6676
4don't6673
5like6265
6go6168
7know6167
8all6065
9just5970
10say5155
11okay4652
12I'm4553
13that's4145
14read4142
15said3842
16good3747
17get3640
18got3637
19big3435
20see3337

Phrase re-read the cat the cat

#ExampleChildrenTimes
1it is it is3443
2a lot a lot1721
3in the in the1618
4I am I am1617
5this is this is1617
6a dog a dog1414
7do not do not1215
8oh no oh no1014
9there is there is1010
10at the at the912
11in a in a911
12a cat a cat910
13on the on the910
14the dog the dog99
15it has it has89
16of the of the88
17I have I have79
18a big a big77
19has a has a77
20it is not it is not77

Extra small word

#ExampleChildrenTimes
1a262362
2the253339
3I195269
4it195250
5and178228
6is171208
7you149168
8that135155
9to127150
10no123150
11what122145
12in113128
13this103114
14not101114
15at98113
16me89100
17on8698
18but8698
19of8293
20he8092

Filler um, uh, er

#ExampleChildrenTimes
1uh7401,415
2oh6661,099
3um392591
4ah285491
5mm198290
6hmm147226
7huh123172
8eh101143
9er5480
10hm3846
11ooh3239
12mhm1011

Word from elsewhere on the page

#ExampleChildrenTimes
1the324527
2a255373
3it155186
4is139165
5to128177
6and119167
7I90115
8in6879
9he6680
10of5879
11not4248
12on4042
13can3638
14that3543
15they3438
16but3135
17you3039
18up3031
19all2833
20are2728

Part-word start b- ball

Ear check: 8 of 11 transcribed starts confirmed, none missed.

#ExampleChildrenTimes
1t- to2223
2c- can2020
3n- not1919
4sh- she1515
5ha- has1314
6b- big1313
7s- see1313
8s- some1313
9th- the1313
10ca- can1212
11fi- fish1212
12I- is1111
13g- get1111
14l- lot1111
15s- she1111
16t- the1111
17s- spot1111
18w- will1111
19a- and1010
20bi- big1010

Self-correction sit → sat

First try → the word it was fixed to.

#ExampleChildrenTimes
1it → is4041
2the → a2429
3is → it2424
4to → the2325
5the → to1919
6it → in1515
7a → the1416
8this → is1414
9and → a1313
10he → his1313
11the → that1213
12that → the1012
13in → it1010
14is → in1010
15the → she1010
16the → they1010
17don't → do99
18it → I89
19the → top89
20to → a88

Stray part-word sound

#ExampleChildrenTimes
1s-95106
2b6980
3e6881
4s6375
5o3541
6x3538
7m3436
8b-3233
9ha-2730
10f-2727
11l2629
12t2627
13p2528
14t-2527
15d2428
16f2325
17k2324
18g-2324
19d-2323
20n1919

Fixed on re-reading

First try → the word it was fixed to.

#ExampleChildrenTimes
1the → a2223
2a → the1010
3I → it55

Sounding out c- a- t

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book wordChildrenTimes
1and88
2a77
3is77
4can77
5the56
6not55
7it44
8she44
9gets44
10help44
11this44
12big33
13cat33
14dog33
15get33
16got33
17sad33
18sit33
19find33
20rest33

Started, then another word

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book wordChildrenTimes
1get55
2spot34
3be33
4dig33
5cave33
6birds33
7big23
8and22
9can22
10elf22
11his22
12pal22
13bees22
14each22
15fine22
16from22
17have22
18hide22
19next22
20seen22

Start after the word dog do-

These forms vary too much to repeat across children, so the list shows the book words they fell on.

#Book wordChildrenTimes
1and66
2the55
3a44
4to33
5are33
6but33
7his33
8she33
9with33
10in22
11is22
12it22
13did22
14lot22
15not22
16set22
17back22
18from22
19gets22
20have22

Pause 0.5–1 s

Each example is the word read after the pause.

#ExampleChildrenTimes
1… the9792,386
2… a9121,921
3… to7051,375
4… and7031,507
5… it6661,214
6… is6391,129
7… in564920
8… I478863
9… he444757
10… on418594
11… of378587
12… they331533
13… that308457
14… she307502
15… but297435
16… can258357
17… not249373
18… with241334
19… up235327
20… no228281

Pause 1 s or more

Each example is the word read after the pause.

#ExampleChildrenTimes
1… the9282,198
2… a7911,528
3… and6441,300
4… it6161,137
5… is577955
6… to561996
7… I521937
8… in492798
9… he464807
10… on342478
11… they339618
12… of334503
13… but301471
14… she300515
15… no276364
16… that265388
17… can257373
18… not251359
19… this244319
20… up232335

Homophone or spacing to → two

Counted as correct: the spelling is the transcriber’s, not the child’s.

#ExampleChildrenTimes
1to → two315404
2an → and101112
3can not → cannot93124
4his → is8192
5to → too7886
6wants → once7582
7a → are6470
8for → four5361
9are → a4147
10want → one4146
11asks → ask4043
12asks → asked3441
13and → an3434
14too → to3334
15is → his3233
16allie → ali2986
17by → bye2930
18indo nesia → indonesia2832
19their → there2831
20sara → sarah2763

Contraction do not → don't

#ExampleChildrenTimes
1it is → it's149210
2I am → I'm4657
3do not → don't3337
4he is → he's2731
5I'm → I am1314
6there is → there's1213
7they are → they're1013
8she is → she's1011
9did not → didn't910
10they're → they are813
11that is → that's89
12does not → doesn't77
13it's → it is67
14it has → it's66
15you are → you're510
16I have → I've55
17can not → can't55
18what is → what's55
19what's → what is55

4,446 children reading passages aloud. Rates count events per 100 book words attempted, from the first to the last book word each recording reached, omissions included. In substitutions the book’s word comes first: ship → shop means the book said ship and the child said shop. A form is listed only if 5 or more children made it, except where the list shows book words. The recordings lean toward harder reading, so rates run higher than in typical reading. Data as of October 9, 2026.

Near misses were the commonest reading errors, at 7.5 per 100 book words: a word close to the target in spelling or sound, or a non-word. Castles, Rastle and Nation (2018) describe the phase this fits: a decoding attempt comes out slightly wrong (“breek” for “break”), and the child can draw on spoken vocabulary to correct it. Different words followed at 4.4, small-word swaps (a for the, he for she) at 3.2 and changed endings at 1.4.

Among disfluencies, word repetition led at 6.5 per 100 words, ahead of extra words or remarks (2.4), phrase re-reads (1.8), extra small words (1.7) and fillers (1.5). Omissions, reversals and voicing swaps were rare, each under 0.5 per 100.

Self-corrections ran at 0.8 per 100, plus 0.2 fixed on re-reading: about one for every 18 errors left standing. That definition is narrow, because a partial attempt that ends on the right word counts here as a part-word start or sounding out.

A words correct per minute score counts disfluencies only as lost time, so it scores about half of these reading miscues as errors.

Two blocks of equal size, errors at about 17 per 100 words and disfluencies at about 18 per 100 words, with only the errors block marked as scored as errors by a fluency test

Part-Word Starts Follow Word Difficulty

Part-word starts (“b- ball”) ran at 1.2 per 100 words. The ear check confirmed 8 of the 11 transcribed starts and missed none, too few to say whether that rate runs high or low. Where they land is the clearer finding.

Part-word starts follow how hard the word is “b- ball” starts per 100 book words

Word length

1 to 3 letters0.68
4 to 5 letters1.89
6 to 7 letters2.61
8 letters or more3.45

How common the word is

The 100 commonest words0.65
Ranks 100 to 1,0001.94
Rarer than rank 1,0003.03
Names and made-up words3.10

Word type

Function word0.48
Content word2.16

Position

First word read in the recording1.07
Any later word1.38

Part-word starts (one onset, then the word) by the book word they lead to. Frequency ranks count how many book pages hold the word. Data as of October 9, 2026.

Part-word starts were about five times as common on words of eight letters or more as on words of one to three, and nearly five times as common on rare words as on the 100 commonest, with names about as high. Content words drew more than four times as many as function words; the first word of a recording, slightly fewer than later words.

That pattern fits a reader working out a hard word from its parts. Part of the length gradient may be detectability, though: a partial start of a one-to-three-letter word is hard to tell apart from the word itself.

Words After Pauses Are Misread More Often

29% of gaps between words lasted half a second or more, and 15% lasted a second or more. A word read after a pause of at least half a second was misread 21% of the time, against 14% after a shorter gap.

This matches Miller and Schwanenflugel (2008), who found that pauses inside sentences were tied closely to word reading skill (r = −.77 in first grade, meaning weaker word readers paused far more) and suggested that long pauses may signal decoding difficulties.

Sound Changes Cluster at Word Ends and in Vowels

Among substitutions within two sounds of the book word, 41% of the sound changes fell at the end of the word, 31% in the middle and 29% at the start. The largest single kind was a vowel swapped in the middle of a word, about a fifth of all changes, spread across many vowel pairs.

The commonest individual changes were consonants dropped or added: /s/ dropped 2,037 times, /z/ dropped 1,857 times, /s/ added 1,606 times and /t/ added 1,182 times. Many /s/ and /z/ changes are probably inflections, a plural or third-person ending read without it, and some may be the transcription missing a quiet final sound. Others may be accent. DIBELS does not penalize consistent dialect or accent (University of Oregon, 2026), and a transcript alone cannot separate the two.

Limits of This Data

These numbers are descriptive and not peer reviewed. They come from children using one reading app with its own books, not a representative sample. Recordings with unclear audio were excluded, which may hide the weakest readers.

Single-word pages were excluded too: children often repeated the word there, which looks like retrying rather than reading. No child’s sentence appears on this page; every example is a word-level form counted across children.

If you work on oral reading errors, every category and example is in the charts above, and we would welcome comparisons with hand-coded samples.


References

Castles, A., Rastle, K., & Nation, K. (2018). Ending the reading wars: Reading acquisition from novice to expert. Psychological Science in the Public Interest, 19(1), 5–51. https://doi.org/10.1177/1529100618772271

Miller, J., & Schwanenflugel, P. J. (2008). A longitudinal study of the development of reading prosody as a dimension of oral reading fluency in early elementary school children. Reading Research Quarterly, 43(4), 336–354. https://doi.org/10.1598/RRQ.43.4.2

Nguyen, T. Q., Del Tufo, S. N., & Cutting, L. E. (2020). Readers recruit executive functions to self-correct miscues during oral reading fluency. Scientific Studies of Reading, 24(6), 462–483. https://doi.org/10.1080/10888438.2020.1720025

University of Oregon. (2026). 8th edition of Dynamic Indicators of Basic Early Literacy Skills (DIBELS): Administration and scoring guide (2026 ed.). https://dibels.uoregon.edu

Frequently Asked Questions

What are reading errors?

A reading error is a departure from the text that leaves a word wrong: a substitution (a near miss such as “shop” for “ship”, a different word, a small-word swap or a changed ending), an omission, or a reversal. In our data from 4,446 children reading aloud, near misses were the most frequent, at 7.5 per 100 book words, followed by different words at 4.4. Disfluencies such as repetitions and part-word starts are usually counted separately, because the word still ends up right.

What is not counted as an error in oral reading fluency scoring?

Under DIBELS 8 rules, repeated words and inserted words are ignored, a self-correction within three seconds is scored as correct, and pronunciation that reflects a consistent accent or dialect is not penalized (University of Oregon, 2026). Hesitations over three seconds, omissions and substitutions are errors. So most disfluencies, including word repeats at 6.5 per 100 words in our data, are never counted as errors in a words correct per minute score, though they cost time.

What is miscue analysis in reading?

Miscue analysis records each place a reader departs from the text, called a miscue, and asks what it reveals about how the reader processed the word. It grew from Kenneth Goodman’s view of reading as a “psycholinguistic guessing game”, which Castles, Rastle and Nation (2018) set against the evidence that skilled reading rests on decoding. Classifying miscues by how close they are in sound and spelling keeps the useful part of the method.

How do you assess reading aloud?

The usual tools are a running record or a timed oral reading fluency probe, scored by a trained listener for accuracy and words correct per minute. Both are hand-scored, which keeps samples small and drops repetitions and self-corrections. The alternative used here is a speech recognition system that keeps partial words, repetitions and fillers, aligned word by word against the book text, with a sample checked by ear.

Download Bookbot now to
GET STARTED