Sources: the research Ladderbug is built on

Reading: method and sequence

Source	What we took from it
Gough & Tunmer, the Simple View of Reading (1986)	Decoding × language comprehension. Why we teach both and measure them separately.
Ehri, phases of word reading development	Why orthographic mapping, and not memorising word shapes, is what makes a word instant.
Scarborough, the reading rope (2001)	Why word recognition and language comprehension are braided rather than sequential.
National Reading Panel (2000)	Systematic phonics outperforms incidental phonics. The basis of the sequence.
UFLI Foundations and LETRS ordering principles	Continuous sounds before stops, confusable pairs separated, one vowel at a time, every new grapheme used in real words immediately.
Three cueing / MSV, and the case against it	Why no picture ever appears beside a word a child has to decode.

Source	What we took from it
Learning Without Tears, print teaching order	Capitals first, grouped by motor pattern rather than alphabetically.
Santangelo & Graham (2015), meta-analysis of handwriting instruction	Explicit instruction improves legibility and fluency, and frees capacity for composition. Modelling plus self-evaluation beats arrows and dots alone.
Longcamp et al. (2005, 2006)	Self-produced handwriting beats typing for letter recognition and specifically for remembering orientation — the mechanism by which forming letters reduces reversals.
The OT Toolbox, on tracing	Tracing alone lets a child start anywhere, which is why it does not transfer.

Source	What we took from it
Common Core progressions, K–5	The major work of each grade, and the order within it.
Cognitively Guided Instruction problem types	Why start-unknown and compare problems are sampled deliberately rather than left to chance.
Concrete–Pictorial–Abstract (Singapore)	Why ten-frames and number bonds come before symbols.

Source	What we took from it
Melby-Lervåg & Hulme (2013); Melby-Lervåg, Redick & Hulme (2016)	Working-memory training does not produce far transfer. Why we make no memory claims.
Simons et al. (2016), Psychological Science in the Public Interest	Review of the commercial brain-training literature.
Aksayli, Sala & Gobet (2019); Sala & Gobet (2020)	Cogmed specifically, and typically developing children: near-zero far transfer.
FTC v. Lumos Labs (2016)	The $2m settlement over claims that brain games improve school and work performance.
Gathercole et al. (2004)	Digit-span norms by age, used to size the memory load in every game.

Source	What we took from it
Hasbrouck & Tindal (2017), oral reading fluency norms	The words-per-minute figures quoted on every age page are the 50th percentile.
Biemiller & Slonim (2001), vocabulary	Root-word counts. Total-word figures run two to three times higher and are much less reliable.

Our mastery rule. Requiring accuracy, speed and stability across two days is a reasonable reading of the fluency literature, but no study tested that exact combination. It is a design judgement, not a finding.
Tracing tolerances. The corridor widths and coverage thresholds by age are derived from developmental norms for letter formation, not measured against a sample of children's traces. We expect to tune them.
Our games. We claim they practise phonics and number facts under memory load. We do not claim, and the evidence does not support, that they expand memory.

If we have misread a source or missed one that contradicts us, we would genuinely rather know than be right. There is a contact link in the footer.