Dating the Undated (Academic Word List register)

Abstract

A considerable body of American radio broadcasting survives as sound recordings, and the surviving material circulates among private collectors as digital audio files. A significant proportion of the surviving material carries no broadcast date, and in many instances the only documentation accompanying a recording is the file name. File names are inherently unstable: file names are rewritten at each transfer, shortened by the operating systems responsible for storage, and, where the text can no longer be read, replaced by a cataloger's estimate. The date is therefore lost, and a significant share of the surviving dates are demonstrably erroneous.

The present paper specifies a method for establishing the true broadcast date of an undated recording. The primary evidence is the newspaper radio log, the day-by-day schedule of radio programs printed in American daily newspapers, and the method derives authority from the comparison of newspaper radio logs published independently in several cities. Trade press listings and evidence internal to the recording itself supplement the newspaper radio log.

The paper reports the results obtained through the application of the method at The Archive of American Radio, a published catalog of 259,915 broadcast records distributed across 6,114 programs. Twenty-eight percent of the 259,915 broadcast records remain undated.

The paper further proposes a framework for making the strength of the evidence behind each date explicit, whereby a date confirmed by several independent contemporary listings is no longer recorded in the same form as an estimate, or in the same form as a date transmitted from another collection and never subsequently validated.

A single case demonstrates the method in operation. Six independent contemporary listings displace the recording's accepted date by almost seven years, and the day of the week cannot resolve the question, because both candidate dates fall on a Sunday and the program was broadcast on Sundays.

The paper also reports a single significant limitation. The number of listings available for a broadcast increases sharply within the sample analyzed, with the year across the 1930s, a trend measured as a correlation of 0.899 across fourteen records. The evidence is not uniformly distributed across the period, and any dating framework must indicate the variation in evidence rather than obscure the variation behind a uniform claim of fact.

Keywords: old-time radio, discography, broadcast dating, newspaper radio logs, provenance, audio archives, metadata

---

1. The problem: a body of evidence whose dates are lost

American radio survives principally because private collectors retained the recordings. The programs were recorded on transcription discs and subsequently on tape, the copies were themselves copied, the copies were traded, and the traded copies were eventually converted into digital files. The sequence of copying and trading continued for approximately seventy years, and throughout seven decades no library or archive imposed a uniform convention on the documentation accompanying each recording. The result is a large body of historically significant material carrying almost no reliable documentation.

The most significant failure is the loss of the date. An undated recording is significantly weaker as evidence, because an undated recording cannot be assessed against the news of a particular week, linked to an advertiser's campaign, or located within the run of the program responsible for generating the recording. Nor can an undated recording demonstrate what a network in fact transmitted on a given evening. Radio was live and scheduled, and then radio was gone; the date is therefore not an attribute among many, but the attribute converting a sound file into a historical document.

Dates are lost through several distinct mechanisms, and separating the mechanisms is useful, because each mechanism requires a different response.

File names break down. Within trading networks the file name carries the documentation. Participants revise the name at each transfer to conform to local convention, operating systems reduce names exceeding a length limit, and characters illegal on a given platform are removed by the person transferring the file. Cataloging software alters the remaining characters. A date written as xx-xx-xx may survive a single transfer; a date written as xx-xx-xx does not survive three.

Erroneous dates are transmitted forward. Where a date is present the date has more often been transmitted than validated. A collector assigns a plausible date, and every subsequent recipient of the file copies the assigned date forward. The assigned date acquires an appearance of authority in direct proportion to the number of repetitions, and no mechanism within a trading network ever eliminates an erroneous date.

Programs are dated; recordings are not. Many holdings record only the years across which the program ran, for example 1952 to 1961. A range of years is a fact about the program, and a range of years is not a fact about the recording.

Estimates are stored as facts. When a cataloger estimates a date, the estimate is written into the same field, in the same format, as a validated date. Nothing marks the difference, and the difference is therefore lost at the first transfer.

The present paper treats the storage of estimates as facts as the principal problem, because a careful method can correct the storage of estimates as facts. The Association for Recorded Sound Collections states the principle explicitly in the association's Guidelines for Discographies: "estimates are preferable to omitting the information. Discographers should explain the rationale behind estimates, and clearly distinguish them from 'solid' data."¹ The requirement is not conclusiveness in every date; the requirement is the reader's capacity to identify which dates are secure, and on what grounds.

1.1 What the existing reference works do, and what the reference works cannot do

The field already possesses authoritative reference works. John Dunning's On the Air: The Encyclopedia of Old-Time Radio documents approximately 1,500 programs, together with the broadcast histories, the scheduled times, the networks, the sponsors, and the principal contributors of the 1,500 programs.² The Handbook of Old-Time Radio, by Jon Swartz and Robert Reinehr, serves the collecting community in a comparable capacity.³ Dunning and Swartz and Reinehr together establish the record at the level of the program, and Dunning and Swartz and Reinehr establish the record authoritatively. The method specified here depends on the work of Dunning and of Swartz and Reinehr: step 2 of the procedure in section 4 consults precisely the encyclopedic category of source in order to obtain a program's scheduled day and hour.

Dunning and Swartz and Reinehr do not, however, date an individual surviving recording, and the reference works were never designed to do so. The subject of Dunning and of Swartz and Reinehr is the program and the program's run, whereas the problem addressed here is the individual item — a single sound file, currently in distribution, whose position within the program's run is unknown. A reference work can indicate to a cataloger the status of Greatest Story Ever Told as a Sunday program broadcast in 1948 and again in 1955, but a reference work cannot specify which Sunday generated the recording in hand. The program-level record analyzed in section 5.4 is equally consistent with the correct date and with the erroneous date.

The limitation of the reference works is therefore structural, and the limitation does not constitute a failure of the existing literature. Dating an individual item requires evidence about the individual item, and for the classic period of American radio the only published record operating at the level of the individual item, day by day, is the newspaper radio log. The present paper argues for treating the newspaper radio log as the primary source for dating undated recordings, and for making explicit in each record which listings were consulted and what the consulted listings establish.

2. The collection

The Archive of American Radio is a public, not-for-profit catalog of American broadcast recordings. The Archive of American Radio published, in September 2026, 259,915 discrete broadcast records distributed across 6,114 programs.⁴ Every broadcast record occupies a persistent address of the form /series/{program}/episodes/{item}, and, where the evidence permits, each record also carries a formatted citation, a set of newspaper listings, and a set of trade press listings.

The collection was not assembled at a single time. The Archive of American Radio aggregates material from the principal old-time radio distribution channels, and The Archive of American Radio consequently exhibits every dating failure specified above, at full strength. A significant proportion of the cataloging effort consists of replacing a transmitted, unvalidated date with a date established against contemporary published evidence.

Two properties of the published catalog can be measured directly, and both are significant for the method.

The available evidence is nearly complete for the classic period. The archive holds sixty records, selected at random, whose item name carries a date between 1930 and 1959. Fifty-nine of the sixty records — 98 percent — carry at least one newspaper or trade press page image.⁵ The evidence required is nearly always present for the years in which the American newspaper radio log was a daily feature, the evidence the method requires is almost invariably available.

Earlier dates persist alongside corrected dates. Of the 259,915 records, 44,566 carry a date in the item name, in the form YYYY-MM-DD, transmitted from the file name of a source collection. Two records in a random sample of forty date-carrying names taken from the whole collection — 5 percent — carry a name date inconsistent with the broadcast date the record itself specifies. One record in sixty was inconsistent within the 1930 to 1959 interval.⁶ Both samples are small, and the proportions should be read only as approximate. Taken together the two samples indicate a residue of one to two thousand records still carrying an earlier date in the name after the catalog has corrected the broadcast date, and section 5.4 analyzes a single inconsistent record in detail.

More than a quarter of the collection is undated. In a random sample of 150 records taken from the whole catalog, 108 — 72 percent — carry a broadcast date, and forty-two — 28 percent — carry none.⁷ A 28 percent proportion across 259,915 records indicates approximately seventy-two thousand items held without a date.

The correspondence between dating and evidence within the sample of 150 records is the more useful result. 101 of the 108 dated records — 93.5 percent — carry at least one page image, whereas of the 42 undated records, two carry a page image. Undated records are not, as a category, records in which evidence was sought and not found; undated records are records no cataloger has yet examined. Dating a record and assembling the record's evidence, in The Archive of American Radio, constitute very nearly the same operation. The forty records in the sample carrying neither a date nor any evidence therefore represent work still to be undertaken, and not a limit on the reach of the method.

2.1 The same broadcast, held more than once

The published catalog exhibits an analogous problem. The archive holds three discrete records for Winston Churchill's speech on Anglo-American unity, delivered at Harvard on 6 September 1943. Two of the three records carry the date Monday, 6 September 1943, together with six page images each, whereas the third record carries neither a date nor any supporting evidence.⁸

Duplication of a single broadcast is normal in a collection aggregated from trading networks. The same broadcast arrives by several routes, and each copy carries a different volume of supporting documentation, so resolving the copies against each other is itself a dating problem. The undated Churchill record is an impoverished record whose answer The Archive of American Radio already holds, fully validated, in the two Churchill records cataloged separately. Resolving a duplicated group of records is among the most productive tasks available to a cataloger, precisely because the underlying evidence has already been obtained.

3. Kinds of evidence

Three categories of evidence are capable of establishing the date of a surviving broadcast recording. The three categories differ in frequency of survival, in precision, and in the characteristic error each category invites. A method treating the three categories as equivalent will generate erroneous conclusions carrying an appearance of certainty.

3.1 Newspaper radio logs

The newspaper radio log constitutes the primary category. The major American daily newspapers printed, from the early 1930s onward, a day-by-day radio schedule, designated the radio log. The radio log specified the station, the hour, and the program name for the subsequent broadcast day. Newspaper radio logs approximate, more closely than any other published series, a complete contemporary record of American broadcast transmission.

Four newspapers sustain the method: the New York Times, the Washington Post, the Chicago Daily Tribune, and the Los Angeles Times. The value of the four newspapers derives less from the authority of any individual newspaper than from the mutual independence of the four newspapers. Each newspaper compiled a radio log from information supplied separately by networks and stations, for a distinct city within a distinct time zone. A program name and hour printed in all four newspapers is corroborated in a manner no individual listing can achieve.

Newspaper radio logs exhibit three weaknesses, and each weakness requires an explicit response. First, a radio log documents the schedule — the transmission a network intended — and not the broadcast. A news bulletin displacing the scheduled program separates the radio log and the broadcast diverge. Second, the radio log went to press before the broadcast day commenced, so the radio log reflects an earlier deadline. Third, newspaper radio logs were printed in very small type within narrow columns. Automated text recognition performs poorly on small type, so transcription errors enter the evidence readily.

3.2 Trade magazines

Broadcasting, the weekly periodical of the American broadcasting industry, together with comparable weekly periodicals, documented schedule alterations, advertiser assignments, network line-ups, and special events. Trade periodicals supply a category of detail the daily newspapers never attempted to supply. Trade coverage appears weekly rather than daily, and trade coverage therefore addresses a smaller proportion of individual dates. Trade coverage is, however, substantially richer in precisely the detail distinguishing a single broadcast from an adjacent broadcast within a program's run. Trade coverage constitutes the natural secondary source wherever a newspaper radio log establishes the transmission of a program without establishing which episode was transmitted.

3.3 Evidence inside the recording

A recording contains internal information capable of supporting a date. A reference to the news establishes an earliest possible date. An advertiser's copy, together with the campaign identified in the copy, can be located within a commercial season. The announcer's name, the network's formal wording, and the wording of station identification all altered over time, and every alteration is datable. A broadcast occasionally names a current event, reporting an election result, a war report, or the outcome of a sporting contest. A named current event frequently locates the recording to the exact day from the recording's own content, and no external listing is required.

Evidence internal to a recording is the only category speaking to the broadcast rather than to the schedule. Evidence internal to a recording therefore adjudicates the question wherever a newspaper radio log and a recording diverge. Evidence internal to a recording is also the category most readily confounded by a subsequent repeat, a subsequent copy, or an Armed Forces Radio Service reissue of the same program.

4. The method

The method is one sentence. Ignore the date on the file, work out roughly which week the recording is from, look that week up in the newspaper radio listings, and count how many of the four newspapers list it. The count is the strength of the date.

The rest of this section explains why each step is needed.

Ignore the date on the file. A date in a file name was written by someone the cataloguer cannot question. Beginning from that date changes the shape of the work: the question stops being which night the broadcast went out, and becomes whether the handed-down date can be kept. A cataloguer who asks the second question will nearly always find an answer, because a plausible date carries a plausible story with it. Write the date down, record where it came from, and set it aside until the end.

Work out roughly when the recording was made. Listen to the recording. A war report, an election result, a sports score, or a joke about the news will often fix the week. Advertising helps, because campaigns ran in seasons. The announcer helps, and so does the wording of the station identification, which changed across the years. None of this yields a single day. It yields a window, and a window is enough to begin searching.

Look that week up in the newspaper radio listings. Every large American daily printed the next day's radio schedule: station, time, and programme. Four papers carry the method — the New York Times, the Washington Post, the Chicago Daily Tribune, and the Los Angeles Times. Read the listings for each candidate date in each of the four papers.

Count the papers that list the programme at the same hour. The count is the answer, and the count becomes the mark on the record:

• A — two or more newspapers list the broadcast
• B — one newspaper lists the broadcast
• C — no newspaper lists the broadcast, but the recording and the programme's known schedule point to the same date
• D — an estimate, given as a range, with the reasoning written down
• E — the date came with the file and nobody has checked it

Counting works because the four papers are independent. Each built its listings in its own city and its own time zone, from copy the networks supplied to each paper separately. Four papers in agreement are four compilers who never spoke to each other reaching the same result. That independence is why two listings, and not one, is the threshold for Mark A.

Mark E carries the argument. A catalogue with no way to record "nobody has checked this" will turn a stranger's guess into a fact, and will do so silently. Section 5.4 is exactly that case: 13 February 1955 never had evidence behind it. It had repetition behind it.

Two cautions. Two papers printing the same network schedule are one source and not two, because the network wrote the schedule once. And a recording that fits a listing may still be a later repeat, so check whether the advertising has been cut out, which is what the Armed Forces Radio Service did when it reissued a programme.

5. Examples

5.1 A broadcast supported by abundant evidence: 1 September 1939

The BBC announcer Alvar Lidell reported the German invasion of Poland. The Archive of American Radio holds the Alvar Lidell report as a recording of forty-five seconds.⁹ The catalog record for the Alvar Lidell report carries ten separate page images. The ten page images comprise newspaper radio logs from the Chicago Daily Tribune, the New York Times, and the Los Angeles Times, together with two distinct listings from the Washington Post. Every listing is dated Friday, 1 September 1939. The corresponding issue of Broadcasting completes the set.¹⁰

The Alvar Lidell date constitutes a Mark A date by a wide margin. The Alvar Lidell record demonstrates the maximum the method is able to achieve. A broadcast of global significance, transmitted on a day of global significance, is listed universally, and confirmation of the date is correspondingly straightforward.

5.2 The same day, a different recording

The Archive of American Radio catalogs Ignacy Jan Paderewski's speech concerning the impending German invasion separately, on the identical day. The Paderewski record rests on the same aggregate of evidence: a Chicago Daily Tribune radio log for Friday, 1 September 1939, together with the corresponding issue of Broadcasting.¹¹ The Alvar Lidell record and the Paderewski record are instructive in combination for what the two records demonstrate about method. A single aggregate of listings for a single day is capable of supporting several distinct broadcast records. The unit of evidence is the printed page, and not the recording.

5.3 A broadcast supported by sparse evidence: 1932

Consider, by contrast, a recording from 1932. The record for Franz von Papen's speech as chancellor, delivered on 13 August 1932, carries three page images, and the three page images derive from the New York Times and the Washington Post alone.¹² The Chicago Daily Tribune listing and the Los Angeles Times listing supporting the 1939 records are absent. The von Papen date is supported, but the corroboration behind the von Papen date is substantially weaker. A framework reporting the Alvar Lidell record and the von Papen record in an identical form would misrepresent the von Papen record.

5.4 A case correcting a transmitted date by seven years

The value of the method resides in correction, and not in the confirmation of an already accepted answer. A clear instance is the Greatest Story Ever Told episode "They Did Not Receive Him."

The recording circulates under a file name carrying the date 13 February 1955. The 1955 date also survives in the item name within The Archive of American Radio, transmitted from the source collection. The catalog record, however, specifies the broadcast date as Sunday, 29 February 1948 — an error of almost seven years in the traded date.¹³

Six independent contemporary listings support the 1948 date. The six listings comprise newspaper radio logs in the New York Times, the Washington Post, the Los Angeles Times, and the Chicago Daily Tribune. A separate "Radio Highlights" item in the Chicago Daily Tribune constitutes the fifth listing, and the corresponding issue of Broadcasting constitutes the sixth listing. All six listings are dated Sunday, 29 February 1948.¹⁴ The date for the episode is therefore a Mark A date.

Two features render the Greatest Story Ever Told case worth extended analysis.

First, the day of the week cannot resolve the case. Greatest Story Ever Told was a Sunday program, and both candidate dates fall on a Sunday. 13 February 1955 and 29 February 1948 were both Sundays. A method resting on a program's established weekday would have accepted the transmitted date without difficulty. Only the six supporting listings discriminate between the two candidate dates. The Greatest Story Ever Told case therefore constitutes the argument for requiring published evidence in place of inference from a schedule.

Second, the correct date is a leap day. 29 February 1948 exists only because 1948 was a leap year. Leap days are an established source of transmitted error. A cataloger reconstructing a date from a damaged or partial file name will ordinarily convert an anomalous 02-29 into a date of conventional appearance. The error within the Greatest Story Ever Told record runs in exactly the direction a leap day predicts.

The citation within The Archive of American Radio already resolves to an address free of the erroneous date, so the corrected date is the date the catalog states and cites. The 1955 text survives only within the item name — an instance of the file name problem specified in section 1, now reproduced inside a catalog having otherwise corrected the broadcast date.

6. Evidence density across the period

The three examples analyzed above suggest a distribution, and the distribution is measurable. The Archive of American Radio holds fourteen cataloged broadcasts drawn from a single news program, covering 1932 to 1939. The number of distinct page images attached to a record, across the fourteen broadcasts, increases strongly with the year of the broadcast, and the correlation is 0.899 across fourteen records.¹⁵ The mean number of page images by year is set out below.

Two explanations are available, and the measurements are not capable of discriminating between the two explanations. The first explanation is that newspaper radio logs genuinely expanded across the 1930s, as radio became a mass medium and the radio log developed from a novelty into an established daily feature. The second explanation is that the collection of newspaper page images by The Archive of American Radio is simply more complete for the later 1930s. The distribution describes the archive rather than the period under the second explanation.

The distinction between the two explanations is significant for the interpretation of the marks. A Mark B date from 1932 is, under the first explanation, as strong as the evidence surviving from 1932 permits. A Mark B date from 1932 is, under the second explanation, merely a record The Archive of American Radio has not finished processing. Discriminating between the two explanations would require a newspaper index independent of the archive's own holdings, and no independent index has been consulted.

The finding is limited in a further respect. The finding rests on fourteen records, all drawn from a single program of international news, and international news is precisely the category of program most likely to appear in every city's radio log. A general claim about the whole collection would require a sample distributed across program categories and networks.

7. Limits

A schedule is not a broadcast. Every date the method derives from a newspaper establishes the scheduled transmission. Programs were displaced, above all during news events, and news events are precisely the occasions on which the collection is richest and the inference weakest.

Repeats and reissues. A recording consistent with a listing may constitute an Armed Forces Radio Service reissue, a regional repeat, or a subsequent copy of the same program. Only evidence internal to the recording is capable of distinguishing a reissue from an original transmission, and evidence internal to the recording is not invariably present.

Automated transcription errors. Newspaper radio logs were printed in dense, small type, and automated text recognition performs poorly on dense, small type. An error within the evidence passes into the catalog without any indication.

Small samples. The finding in section 6 rests on fourteen records drawn from a single program. The collection-wide proportions in section 2 derive from the published catalog, and completing the proportions requires the cataloging history held internally by The Archive of American Radio.

Reproduction of the evidence. For the classic period of American radio, the newspaper pages supplying the evidence remain under copyright. The present paper cites listings by newspaper, date, and category of listing, and the present paper does not reproduce the newspaper pages.¹⁶

8. Publishing the reasoning, not only the answer

A mark for the strength of evidence resolves nothing while the mark resides only in a cataloger's private notes. The failure specified in section 1 is precisely the flattening of careful decisions into bare assertions as recordings pass from collector to collector. A mark must be published with the record to survive the trading networks destined to carry American radio recordings forward, the mark must be published alongside the record and attached to the record durably.

Three requirements follow, and The Archive of American Radio implements all three requirements.

The citation belongs to the record, and not to a report. Each broadcast record carries a formatted citation and a list of supporting listings beside the date, at the identical web address. A reader reaching a broadcast record reaches the evidence for the broadcast record. Publishing the citation beside the date is the mechanism rendering the undated copy in section 2.1 immediately visible. Two records state a date supported by six page images each, whereas a third record states nothing, and any reader is able to perceive the difference.

The item name must not carry a competing claim. Section 5.4 constitutes a warning. The catalog holds the corrected date, cites six listings for the corrected date, and directs the citation at an address free of the erroneous date. The item name nevertheless still reads 1955-02-13, and the item name is the text an automated agent reading web addresses will aggregate. An item name containing a date advances a claim, irrespective of the cataloger's intention. The estimate in section 2 indicates a residue of one to two thousand records in an identical state. Dates belong in fields, and not in names.

Absent information must be recordable. Twenty-eight percent of the collection carries no date, and the catalog states the absence rather than supplying a plausible year. The explicit statement of absence is the requirement the Association for Recorded Sound Collections imposes on discographers in the passage quoted in section 1. The alternative, for a date lacking support, is not an empty space awaiting completion. The alternative is an explicit statement that the question remains open, and Mark E carries the statement. The proportion of two in forty-two — undated records almost never carrying page images — is the measurement rendering Mark E useful, because the proportion identifies undated records as work awaiting completion.

9. Conclusion

The dating problem in American radio is not principally a problem of absent evidence, because the evidence largely survives. The evidence resides in the newspaper radio logs of four major daily newspapers and in the trade press, and the greater proportion of the evidence is now available in digital form. The problem is that the documentation accompanying American radio recordings possesses no mechanism for expressing the difference between a date validated against contemporary published evidence and a date transmitted without evidence. A validated date and a transmitted date are consequently no longer distinguishable from each other.

The remedy proposed here is straightforward. Attach a mark for the strength of evidence to every date. Maintain a visible category for dates transmitted and never validated. Cite the exact published listing behind every confirmed date. Applied across 259,915 records, the marking framework converts a trading collection into a collection scholars are able to cite.

---

Notes

1. Association for Recorded Sound Collections, Guidelines for Discographies. ↩
2. John Dunning, On the Air: The Encyclopedia of Old-Time Radio (Oxford University Press, 1998). ↩
3. Jon D. Swartz and Robert C. Reinehr, Handbook of Old-Time Radio: A Comprehensive Guide to Golden Age Radio Listening and Collecting (Scarecrow Press, 1993). ↩
4. Counted from the published sitemap of radioindex.org, 1 September 2026: 259,915 episode addresses across 123 sitemap files, and 6,114 series addresses. One file (episodes-96.xml) returns a valid but empty list; it is counted among the files and excluded from the record count. ↩
5. Random sample of sixty cataloged broadcasts drawn from the 40,727 item names carrying a date between 1930 and 1959; retrieved 1 September 2026. Fifty-nine carried one or more page images. ↩
6. Random samples of forty records from all 44,566 date-carrying item names, and sixty from the 1930 to 1959 group; retrieved 1 September 2026. A disagreement means the date in the item name differs from the broadcast date the catalog record states. ↩
7. Random sample of 150 records from the whole catalog; retrieved 1 September 2026. ↩
8. "Winston Churchill on Anglo-American Unity," Radio News; and "Anglo-American Unity" (two records), Winston Churchill Speeches and Radio Broadcasts. Radio Index, retrieved 1 September 2026. The first two carry the date Monday, 6 September 1943 and six page images each; the third carries neither date nor image. ↩
9. "Alvar Liddell Reports the German Invasion of Poland." WWII News. Radio broadcast, 1 September 1939. Radio Index. https://www.radioindex.org/series/wwii-news/episodes/alvar-liddell-reports-the-german-invasion-of-poland. ↩
10. Radio logs, Chicago Daily Tribune, New York Times, Los Angeles Times, and Washington Post, all Friday, 1 September 1939; and Broadcasting, 1 September 1939.¹⁷ ↩
11. "Ignacy Jan Paderewski on the Nazi Invasion Looming over Poland." WWII News. Radio broadcast, 1 September 1939. Radio Index. https://www.radioindex.org/series/wwii-news/episodes/ignacy-jan-paderewski-on-the-nazi-invasion-looming-over-poland. ↩
12. "Chancellor Franz von Papen." WWII News. Radio broadcast, 13 August 1932. Radio Index. Supporting listings: New York Times and Washington Post radio logs, Saturday, 13 August 1932. ↩
13. "They Did Not Receive Him." Greatest Story Ever Told. Radio broadcast, 29 February 1948. Radio Index. https://www.radioindex.org/series/greatest-story-ever-told/episodes/they-did-not-receive-him. The item name keeps the wrong inherited date 1955-02-13. ↩
14. Radio logs, New York Times, Washington Post, Los Angeles Times, and Chicago Daily Tribune; "Radio Highlights," Chicago Daily Tribune; and Broadcasting — all Sunday, 29 February 1948. ↩
15. Sample of fourteen cataloged broadcasts from WWII News, retrieved 1 September 2026; page-image counts taken from each record's newspaper and periodical sections. ↩
16. On the copyright of the recordings themselves: United States law, at 17 U.S.C. section 1401, brought sound recordings made before 1972 under federal copyright. Recordings first published between 1926 and 1946 are protected for 100 years, and those from 1947 to 1956 for 110 years. Unpublished recordings made before 1972 are protected until 15 February 2067. There is no general rule that material from before 1972 is free of copyright. ↩
17. Following the Chicago Manual of Style, 18th edition, page numbers are left out of newspaper citations throughout this paper. A newspaper may print several editions in one day, and items move or disappear between them. The newspaper, the city, the date, and the kind of listing identify the item; the page number does not do so reliably. ↩

---

Discography

Following the Chicago Manual of Style, 18th edition, sound recordings are listed here rather than in a bibliography. The Archive of American Radio holds the recordings listed below, and each recording is cited by the recording's permanent web address. The dates supplied are the broadcast dates established by the method in section 4, with the strength mark in parentheses.

"Alvar Liddell Reports the German Invasion of Poland." WWII News. Radio broadcast, 1 September 1939 (Mark A). Radio Index. https://www.radioindex.org/series/wwii-news/episodes/alvar-liddell-reports-the-german-invasion-of-poland.

"Anglo-American Unity." Winston Churchill Speeches and Radio Broadcasts. Radio broadcast, 6 September 1943 (Mark A); and a second record of the same broadcast carrying neither a date nor a supporting listing (Mark E). Radio Index.

"Chancellor Franz von Papen." WWII News. Radio broadcast, 13 August 1932 (Mark A). Radio Index.

"Ignacy Jan Paderewski on the Nazi Invasion Looming over Poland." WWII News. Radio broadcast, 1 September 1939 (Mark A). Radio Index. https://www.radioindex.org/series/wwii-news/episodes/ignacy-jan-paderewski-on-the-nazi-invasion-looming-over-poland.

"They Did Not Receive Him." Greatest Story Ever Told. Radio broadcast, 29 February 1948 (Mark A). Radio Index. https://www.radioindex.org/series/greatest-story-ever-told/episodes/they-did-not-receive-him. Circulates under the erroneous inherited date 13 February 1955; see section 5.4.

"Winston Churchill on Anglo-American Unity." Radio News. Radio broadcast, 6 September 1943 (Mark A). Radio Index.

The twenty WWII News records examined for section 6 appear with the corresponding supporting sources in the Appendix, and are not repeated here.

---

Bibliography

Dunning, John. On the Air: The Encyclopedia of Old-Time Radio. Oxford University Press, 1998.

Swartz, Jon D., and Robert C. Reinehr. Handbook of Old-Time Radio: A Comprehensive Guide to Golden Age Radio Listening and Collecting. Scarecrow Press, 1993.

Newspaper radio logs and trade magazine listings are cited in the Notes and are not repeated here, following the Chicago Manual of Style, 18th edition, which leaves newspaper items out of the bibliography.

---

Sources for the discography

The present section specifies the sources the method depends upon, the coverage each source provides, and the points at which the coverage fails, so a reader is able to replicate the sequence of steps described in section 4.

Newspaper radio logs. Four major daily newspapers sustain the method: the New York Times (New York), the Washington Post (Washington), the Chicago Daily Tribune (Chicago), and the Los Angeles Times (Los Angeles). Each newspaper compiled a radio log from information supplied for the newspaper's own city and time zone. The independence of compilation is the reason agreement among the four newspapers constitutes corroboration rather than repetition. Two categories of listing are counted separately where a newspaper printed both: the daily radio log, a complete schedule, and radio highlights, a selected column composed by an editor. The Chicago Daily Tribune record for 29 February 1948, analyzed in section 5.4, supplies a single instance of each category, and the two items count as two listings because the two items were produced independently.

Coverage is not uniform. The number of listings available for a broadcast increases, within the archive's holdings, across the 1930s, from a mean of 3.0 in 1932 to a mean of 7.8 in 1939 within the sample analyzed in section 6. Ninety-eight percent of the cataloged broadcasts sampled for 1930 to 1959 carry at least one page image, whereas across the whole collection the corresponding proportion is 68.7 percent. Almost the entire difference derives from records carrying no date, of which only two in forty-two carry any page image at all.

The provenance of the newspaper images. The catalog presents each newspaper listing as a page image, labeled with the newspaper, the city, the date, and the category of listing, and the catalog does not record the provenance of the scan. The published version of the present paper should name the provenance explicitly. The published version should specify whether each run was consulted in a library newspaper database, on microfilm, or within a scan collection assembled by a third party, and the published version should specify what is established about the completeness of the source. The paper should, where the source constitutes a private collection rather than a library, state the fact plainly, together with what is established about the collection's history and about any rights the collection asserts. A reader cannot replicate a sequence of steps whose starting point is unnamed, and recording the provenance is the single most useful addition the apparatus still requires.

Trade magazines. Broadcasting, the weekly periodical of the industry, is consulted through WorldRadioHistory, which the catalog names and cites as the source of the scan. Trade coverage is weekly rather than daily. Trade coverage supports schedule alterations, advertiser assignments, network line-ups, and special events — the category of detail distinguishing a single broadcast from an adjacent broadcast within a run wherever the daily newspaper radio logs supply only a program name. Broadcasting appears within the sample analyzed for the better documented broadcasts, and is absent for the sparsely documented broadcasts. Trade coverage supplements the daily newspaper radio logs, and trade coverage does not replace the daily newspaper radio logs.

Evidence inside the recordings. The recordings constitute the archive's own holdings. Each recording occupies a persistent address of the form /series/{program}/episodes/{item}, and each recording carries a formal citation. The citation, where the present paper cites evidence internal to a recording, refers to sound audible within the archive's copy.

The points at which coverage fails. Three failures are established, and any reader replicating the present work should anticipate all three failures. Newspaper radio logs document the schedule rather than the broadcast, so a displaced program breaks the correspondence, and displacement occurs most frequently during news events, precisely when the collection is richest. Newspaper radio logs went to press before the broadcast day commenced. And newspaper radio logs were printed in dense, small type, which automated text recognition reads badly, so errors within the evidence pass silently into the catalog. Adding further newspapers corrects none of the three failures, because all four newspapers share all three failures.

---

Appendix: Dated log

The log below covers the WWII News recordings analyzed for the present paper, in order of broadcast date. The log follows the conventional form for a discography: the date, the title, the running time, the number of distinct supporting listings, the supporting listings by short name, and the strength mark from section 4.

Short names: NYT New York Times; WP Washington Post; CDT Chicago Daily Tribune; LAT Los Angeles Times; Bcstg Broadcasting. Both are counted where a newspaper printed a radio log and a separate radio highlights item for the identical date, both items are counted, because the two items were produced independently.

Every record within the log carries Mark A, and each record is supported by two or more independently produced contemporary listings. The uniformity is a property of the program sampled rather than a property of the whole collection. WWII News is international news, the category of program most reliably printed in every city's radio log. A log drawn from local daytime drama would exhibit a very different distribution of marks, and section 6 should be read accordingly.