---
title: "Widrow's own retrospectives describe Lucky's Bell Labs equalizer as an independently developed, different algorithm, with LMS entering Bell Labs' adaptive-modem work only afterward"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://archive.computerhistory.org/resources/access/text/2014/01/102746758-05-01-acc.pdf"
source_author: "Bernard Widrow, interviewed by Shayne Hodge"
source_date: "2013-04-15T00:00:00.000Z"
source_quote: "there's a man named Robert Lucky who is very famous in the annals of the IEEE... And Lucky developed the adaptive equalizer at Bell Labs. He had developed a learning algorithm, which is different from LMS, and, in my opinion, not as good... But somehow LMS got in there and got into their work on the adaptive modems. And it was the work on adaptive modems at Bell Labs that really caused it to get into digital communications."
source_tier: 1
audit_status: "capture-verified — fetched this session with tls: verified; two corroborating Tier-1 Widrow-authored texts read alongside it (2005 IEEE Signal Processing Magazine retrospective, 1990 Widrow & Lehr Proceedings of the IEEE survey); independent cross-model re-check not yet performed. | 2026-07-22 cross-model audit (fable): all three primaries re-extracted and read directly (CHM oral history sha256 4c0111d2…, p.21 quote + '65/'66/'67 + 'That I'm not sure' verbatim; 2005 SPM 'Thinking About Thinking' sha256 c3c26dd2…, p.106 quote verbatim; Widrow & Lehr 1990 sha256 f41e3cc5…, pp.1415–1416 quotes verbatim). Two corrections applied: (1) '[sic]' added — the 2005 article misprints 'M.M. Londhi' for M.M. Sondhi (canceller is Sondhi & Presti, BSTJ 45, Dec. 1966, archive.org bstj45-10-1851); (2) the 1990 'independent path' sentence was being read as Stanford-vs-Bell-Labs framing, but its in-context referent is the neural-net paradigm developments (ART, Hopfield, Boltzmann) — body reworded to rest the 1990 corroboration on the Lucky-applications sentence instead. Core sequencing claim unaffected; verified."
provenance: "Promotion from 10-inbox/raw/2026-07-19-were-luckys-bell-labs-and-widrows-stanford-adaptive.md, 2026-07-21"
origin: "batch"
derived_from: ["20260719-0213-were-luckys-bell-labs"]
date_created: "2026-07-21T00:00:00.000Z"
tags: ["adaptive-filtering","bell-labs","widrow","lucky","lms-algorithm","priority-and-independent-discovery","history-of-technology"]
audits: ["2026-07-22 claude-fable-5"]
---


In a 2013 Computer History Museum oral history, asked how LMS came to be
adopted outside his own lab, [[entity-bernard-widrow|Bernard Widrow]] answered: "there's a man named
Robert Lucky who is very famous in the annals of the IEEE... And Lucky
developed the adaptive equalizer at Bell Labs. He had developed a learning
algorithm, which is different from LMS, and, in my opinion, not as good...
But somehow LMS got in there and got into their work on the adaptive modems.
And it was the work on adaptive modems at Bell Labs that really caused it to
get into digital communications." Asked for a date, Widrow placed it loosely
— "probably '65, '66, '67" — and added, of LMS's actual adoption: "When they
actually started using LMS? That I'm not sure."

Two other Widrow-authored primaries corroborate the same sequencing. His
2005 IEEE Signal Processing Magazine retrospective states: "Use of the LMS
algorithm really took off after the invention of the adaptive equalizer by
R.W. Lucky and the adaptive echo canceller by M.M. Londhi [sic — the
article's misprint for M.M. Sondhi; the canceller is Sondhi & Presti, "A
Self-Adaptive Echo Canceller," BSTJ 45, Dec. 1966] and A.J. Presti at
Bell Telephone Laboratories in the 1960s and 1970s." The 1990 Widrow & Lehr
*Proceedings of the IEEE* survey (already on file behind
[[claim-madaline-rule-i-first-layer-trainable-second-fixed]]) gives the same
sequencing from the Stanford side: it credits "outstanding work by Lucky and
others at Bell Laboratories" with leading to "major commercial applications
of adaptive filters and the LMS algorithm" in high-speed-modem equalization
and echo cancellation — LMS arriving at Bell Labs as an applied import. (The
survey's "adaptive systems research at Stanford traveled an independent
path" sentence, in context, claims independence from the neural-network
developments just listed — ART, [[entity-john-hopfield|Hopfield]] nets, Boltzmann machines — not
from Bell Labs specifically, so it corroborates parallel lines only weakly.)

Read together, all three sources place Lucky's original algorithm as
distinct from LMS at its 1965 origin, with LMS specifically arriving at Bell
Labs as a later graft onto Lucky's independently-invented equalizer
([[claim-lucky-1965-adaptive-equalizer-steepest-descent-transversal-filter]]).
Every source here is Widrow's own account of his side of the story, spanning
1990–2013 — see
[[claim-lucky-widrow-independent-origination-favored-lucky-side-unverified]]
for what this leaves unaddressed.

> [!note] Seek's commentary:
> Widrow calling a rival algorithm "different from LMS, and, in my opinion,
> not as good" is the kind of unguarded aside that's worth more than the
> polished retrospective prose around it — three decades on, in a recorded
> interview, he still bothers to rank it. That's a small tell that he's
> speaking from memory of a real technical distinction, not reciting a
> credit narrative he's rehearsed. It doesn't make him a neutral witness —
> nobody is neutral about their own algorithm's rivals — but it does make
> him a specific one, and specific beats smooth.
> — Seek