Transcription Procedure:
Audio files were first transcribed by an external agency. Given that audio-only transcription of parent-child interaction is challenging and that automated systems often miss important details, subsequent transcription phases were crucial for providing a high quality transcript for language analyses. Therefore, each initial transcription was subjected to a series of additional steps; the transcription work flow is summarized below:
*headphones were used for the first two transcription phases
●	Prepping Phase: 
○	Research Assistants (RAs; who have complete training for independent prepping after reformatting at least two training transcripts and reviewing detailed feedback from lab staff) converted the audio-only transcript to one that structurally resembled a CHAT transcript removing names to de-identify. 
○	RAs made edits to the transcript while watching the parent-child interaction video, improving accuracy for speaker tiers, utterance boundaries, lexical material, and untranscribed sections. This resulted in a prepped transcript.
●	CHAT Phase:
○	A CHAT trained RA or SLP (Speech Language Pathologist; reliable via training standards set by the lab/project protocol including reaching 90% reliability for utterance boundaries and utterance delimiters on at least two training transcripts) watched the parent-child interaction and used the prepped transcript to generate a CHAT transcript by further improving utterance boundaries, lexical material, untranscribed or unintelligible sections, and adding specific CHAT conventions.
○	Each transcript-video set was viewed by at least two team members, always including an SLP.
●	Proofreading Phase:
○	CLAN features including: CHECK, FREQ, and MOR (10.21.25 version) were run on transcripts to identify typos and unrecognized lexical material.
■	Unrecognized words were added to the corpus lexicon in a cut file (detailed below). Where the MOR lexicon has multiple common spellings for a word, we selected one spelling for internal consistency.
○	Batchalign2 was run to identify any CHAT specific typos.
○	RAs then read the transcript (no video viewing) looking for typos, mis-matched speaker tiers, etc. to identify any remaining errors before flagging it as ready for upload to TalkBank.

Transcription Conventions
Time Stamps: 
●	Time stamps and Headers were included for each section corresponding to the parent-child interaction protocol, and GEM Markers (@G) were added.
○	Section names include: Free play/Conversation, Pretend Play Independent, Pretend Play with Parent, Sensory/Cause and Effect Play.
○	Judgement was made as to when the transcribers felt that each specific section truly started, but always after directions were provided by the INV.
●	An end timestamp was added when the play finished and transcription stopped.  
●	The accompanying videos may start before the transcript and/or may end after the transcript based on when the section was judged to have begun.
○	All videos were trimmed following the completion of the parent-child interaction. 
Privacy:
●	Personally identifiable information was removed with video blurring and audio editing as needed.
●	We also used the following anonymization conventions in the transcripts:
○	Use of www with an explanation tier when personal information was produced.
○	Use of Childname, Mothername, Fathername, Investigatorname, Firstname, Schoolname, Petname, etc.

Detailed CHAT Conventions:
Lexicon
●	Project staff built a lexicon using CLAN’s ENG folder and MOR 10.21.25 version with additions specific to this study, including: 

aac	{[scat n]}	n-abbrev
aba	{[scat n]}	n-abbrev
atv	{[scat n]}	n-abbrev
crisscross_applesauce	{[scat co]}	co-rhyme
off+roading	{[scat n][comp n+part]}	n+n+part
parkour	{[scat n]}	n
polyhedron	{[scat n]}	n
sriracha	{[scat n]}	n

Yes/No Spellings
●	mhm (meaning yes)  
●	uhhuh (meaning yes)
●	hmhm (meaning no)
●	hunhunh (meaning no)
●	nahunh (meaning no)
Shortened Words
●	The following shortened/incomplete words were transcribed as full words for readability:
○	cause→because
○	em→them
○	whatcha → what are you or what you
○	dontcha → don’t you
○	lemme → let me
●	The following shortened words were transcribed as they were spoken:
○	wanna 
○	gonna
Proper Nouns & Titles
●	Character names and/or titles were transcribed as proper nouns (e.g. McQueen, Baby)  and received an underscore as a linkage as needed (eg. Peppa_Pig, Spider_Man, Ninja_Turtles, Old_MacDonald).
●	Nicknames and/or terms of endearment were capitalized and thus transcribed as proper nouns  (e.g. Honey, Sweetie, Sweetheart, my Love).
Questions and Recasts
●	Intonation was primarily used to determine if inclusion of a ? utterance delimiter was warranted.
Vocal Actions were transcribed using &= and were not phonetically transcribed.
○	Examples include but are not limited to: laughs, gasps, sighs, grunts, squeals, screams, hums.
○	If these actions (e.g. laughing) occurred along with production of lexical material, the <”> [=! laughing] notation was used.
○	Audible inhalations (i.e. not gasps) were not coded with this notation and instead were coded in a comment tier.
○	Audible inhalations that were solely part of inhaling to blow up a balloon were ignored.
Vocalizations were judged as non-word like or word-like.
○	Non-word vocalizations: vocalization that cannot be clearly characterized as a true word approximation. It may be sensory in nature. It is likely mostly vowel only (V) or a combination of consonant-vowel (CV or VC.) 
■	A phonology tier was added as often as possible, especially for CV or VC.
■	If the speaker had a toy, food/drink, or a pacifier in their mouth, the vocalization was not phonetically transcribed.
○	Word-like vocalization: these can be considered “words” for the purposes of identifying words/utterances. This may contain “novel” jargon or scripted unintelligible language.
■	These may include sounds (syllables of at least one vowel and one consonant or repeated syllables, jargon, different syllables strung together).
■	These were coded as yyy with an accompanying phonology tier.
■	If the word-like vocalization (yyy) is in a multi-word utterance, the phonology tier includes phonetic transcription of yyy and the ellipsis symbol for all other words and/or conventions on the corresponding speaker tier.
Gestures were not the primary focus of the project and therefore were not specifically coded using &= notation.
○	Gestures that seemed to convey communicative intent were included as comment tiers for all participants.
■	Common gesture descriptions include: points, beckons, nods head-yes, shakes head-no, and other descriptions such as turns hands out to side, palms up as if to question.
Animal Sounds were transcribed with @o (onomatopeia) if they were commonplace, recognizable sounds (e.g. moo, woof, neigh, meow). If other other animal sounds were made, they were transcribed using &=imit: “animal” notation.
●	With respect to a *MED tier (talking toy):  If the animal sound was spoken as a word (e.g. moo) it was transcribed as such. If the toy played a cow sound, it was transcribed as &=imit:cow.