The Caledonia Mills case in Nova Scotia became famous for fires, disturbed farm objects and reports of unseen blows. Walter Franklin Prince investigated in March 1922 and published an account which is often reduced to a simple verdict. The original report is more complicated. He strongly favoured human agency for the fires, distinguished other alleged experiences from them and expressly acknowledged that he had not observed the earlier disturbances occurring.
The family’s reports and the investigation
Prince places the main disturbances in December 1921 and January 1922, with the fires forming the final phase. The household was that of Alexander MacDonald and his family near Caledonia Mills in Antigonish County. In his account, numerous small fires occurred from 6 to 12 January. Other reports concerned released cattle and displaced clothing. Reporter Harold Whidden and a detective named Carroll later stayed in the emptied house and described sounds and sensations of being struck. Prince’s original report, printed pages 422–441.
Prince initially wrote to the newspaper reporter for a first-hand statement. He subsequently undertook the investigation with support arranged by W. H. Dennis, proprietor of Halifax newspapers, while insisting on control of his inquiry and freedom to report whatever he concluded. He arrived in Halifax on 4 March. The journey to the house required a long sleigh trip, supplies and arrangements to stay in a deserted building during severe winter weather.
His party occupied the house from the afternoon of 7 March until the morning of 13 March. Prince spent the final night there alone. No new fires or comparable disturbances occurred during that stay, nor did the earlier reported sounds and blows recur. He explicitly warns that non-recurrence by itself would not prove or disprove a proposed cause. His investigation of the fires therefore relied on surviving traces, interviews and reconstruction rather than witnessing an ignition.
Why Prince proposed human agency
Prince inspected the locations and character of the remaining burns. He argued that their heights and accessibility were consistent with a person of the foster daughter Mary Ellen’s stature. He also considered whether the visible parts of rooms, the positions of furniture and movement through the house would have allowed someone to act unseen. These were physical observations followed by an inference about agency. They were not a direct observation of the girl starting a fire.
Some particular findings strengthened his view. He reported a match among burned cloth in a recess and materials in the house which could have contributed to ordinary combustion. His account distinguishes between something being available and proof that it was used. That distinction remains necessary: the presence of a potentially relevant material does not establish its role in every fire.
He also set aside the first fire near the stovepipe because it might have had an ordinary accidental origin. His theory thus did not require all the outbreaks to begin in precisely the same way. A dramatic series can contain different mechanisms, and an investigation must test its proposed explanation against each part rather than assume that one initial cause necessarily continued throughout.
The limits of the accusation
Prince rejected the idea that witnesses had deliberately lied. He argued that honest observers could miss actions during a confusing succession of events or remember their sequence imperfectly. A witness who had suspected the girl and believed he watched her carefully did not persuade him that ordinary human agency was excluded. That was Prince’s judgement of the testimony, not a demonstration that every witness’s account was mistaken.
He also opposed a straightforward moral accusation against Mary Ellen. His proposed explanation included an altered state of consciousness, using the psychological language of his period. He entertained, without adopting a firm conviction, a possible spiritual influence as an alternative speculation. These passages should be read as historical interpretations. They do not establish a modern diagnosis, a confession or a proven finding of deliberate wrongdoing.
This matters when summarising the case. Saying that Prince blamed human hands for the fires accurately conveys a major conclusion. Saying that he caught the girl setting them, or settled every event as a conscious hoax, goes beyond what this report says. His certainty about one inferred physical cause coexists with qualifications about responsibility and with different assessments of the other allegations.
Sounds, blows and automatic writing
Prince treated Whidden and Carroll’s reports of unusual noises and tactile experiences as a separate class. He leaned towards a supernormal interpretation by comparison with experiences in other cases. However, he had not observed these particular events, and he explicitly says he had not been able to interview Carroll. His argument relied substantially on Whidden’s account and on analogies which Prince considered significant.
Those analogies explain his reasoning, but cannot independently verify a sound or an unseen blow. The absence of comparable events during the later stay also left him without the opportunity to test their immediate circumstances. The original report preserves the difference between this testimonial material and the physical burn marks which he could actually examine.
An experiment with automatic writing on 10 March introduced a new episode during the investigation. Prince selected Whidden for the trial. After an initial failure, writing occurred, followed by further unsuccessful attempts. Prince considered the writing a genuine psychological occurrence, while saying that its contents did not then permit a firm decision that they exceeded the writer’s own mind. He discussed a possible connection with the earlier disturbances without claiming that he had established it.
A report caught in its own publicity
Prince’s final pages complain about statements, actions and supposed interviews falsely attributed to him in newspapers. He says that accounts had circulated while the investigating party was isolated and that he preferred people to rely on his signed words. This warning is especially useful because publicity is part of the surviving case, rather than a neutral container for it.
Contemporary reporting in the Morning Oregonian on 16 March carried an Associated Press account of his conclusions; the following day reported local criticism of them. These items show how quickly a complex inquiry became an argument about the witnesses’ credibility. They should not replace the fuller signed report. 16 March 1922; 17 March 1922.
Assessment and geographical limits
UKPX classifies Caledonia Mills as Possibly Explained. Prince provides a concrete argument for ordinary human agency in the fires, with an accidental possibility retained for the first outbreak. He did not witness the earlier ignitions, and his separate interpretations of sounds, tactile experiences and automatic writing do not amount to a demonstrated explanation of the entire case. Neither a haunting nor a specific psychological cause is established here.
The map uses an approximate Caledonia Mills locality reference, GeoNames 5913435, rounded to 45.49, −61.80, in Nova Scotia, Canada. It is not the MacDonald home, a private residential address, a surviving ruin or a visitor destination. “Near Antigonish” in the report title refers to the wider county setting. Coordinate source: GeoNames Canada dataset.
Original sources and related cases
The twenty pages of Prince’s report were read in the original 1922 journal scan, with the opening page and concluding summary visually checked. The lead image reproduces that opening page. It is a source document, not an image of the fires. For separate Nova Scotia and Canadian cases, see Amherst and Esther Cox and the Dagg haunting investigation. Similar reported effects do not by themselves demonstrate a shared cause.