No source exists that names Mary Caroline's parents. The only family lore (per Shari S) is that her children claimed that she was part Cherokee and that she died sometime after the family moved from Arkansas to Oklahoma (if not on the way). In her daughter Lizzie's old cigar box of photos was a tin type of a man in a suit holding a Bible who appears to be Native American; however this may have been a photo that originated with Lizzie's husband's family, and if it has any relationship to anyone is unknown. (Photo now kept by Shari S.)
Various theories for her parents have been established based on likely matches in the 1870 census. It is hoped that DNA testing with segment triangulation will eventually help narrow down her origins.
The theory of descendant Shari S is that Mary Caroline, born around 1861, is the ten-year-old "Carline Pearry" of the 1870 census in Madison County, Arkansas. (Digital image of original document at Ancestry.com: http://search.ancestry.com/cgi-bin/sse.dll?db=1870usfedcen&h=13809658&indiv=try). Hilburn Township was five miles from White River Township, where Mary Caroline was enumerated in 1880.
That household is headed by Mary J. Pearry, who was enumerated alone in 1860 Madison County (Valley Township) as well. The Arkansas 1911 Census of Confederate Veterans, page 60, has an entry for her son David H. Perry. Ancestry.com user HAMPELIPPCHE has transcribed the entry in his tree: "That David H.Perry of Cabanal, Carrol Co, AR was born April 7, 1845 In Fannin Co, GA, the son of David Gipson Perry who was the son of David Gipson Perry. Maiden name of subject's mother was Mary Morton. He states he had some uncles in the Mexican War. Subject was educated in Madison Co, AR, was a Farmer, Democrat, United Baptist and Odd Fellow. He was a PVT. in Shelby Command, Scotts CO, Nichols Regiment and served from 1862 to 1865...."
Madison County Cemeteries (Volume II), published in the summer of 1998 by the Madison County Genealogical and Historical Society, describes birth and death dates for Mary Morton Perry on page 273: 29 August 1809 and 7 May 1879. If these dates are correct, Mary would be close to sixty when Mary Caroline Perry was born. However, both the 1860 and the 1870 census place her birth year around 1820. (That said, the 1850 census for D. G. Perry - page 261a in Polk County, Tennessee - gives a circa 1809 birth year for his wife Mary.) With this in mind, even if "Carline Pearry" of 1870 is Mary Caroline, it's possible that she is a niece or cousin of this family, not a daughter or sister.
Great-great-grandson C. Simonds autosomal test results at Ancestry.com include an extremely high confidence match (predicted 4th cousin) to a descendant of John H. Perry, son of David Gibson Perry and Mary J. Morton as well as a high confidence match (5th-8th cousins) of another descendant of John H. Perry, a close cousin of the first match, thus reducing the possibility that the connection is via another, unknown ancestor.
C. Simonds has a "good" confidence match (5th-8th cousins) to a descendant of Addison Perry, of Madison County, Arkansas. Addison's family lived near the descendants of David Gibson Perry for decades. L. Simonds has an additional DNA match to a descendant of Addison's brother (same confidence level).
Unfortunately, Ancestry.com's DNA test does not show segments, and these matches have not yet uploaded to GEDmatch. Testing at 23andMe and FTDNA has not yet yielded Perry matches with links to Arkansas, although this may be due to the scarcity of family trees or surnames listed compared to Ancestry.com, and the sheer number of matches making one-on-one communication a long process. Until segment triangulation takes place, Mary Caroline Perry's relationship to the Madison County Perry families is speculative, albeit with compelling evidence.
It should also be noted that so far Mary Caroline's descendants through daughter Lizzie's daughter Pearl have recorded no DNA matches to relatives of her husband, Lewis Marion White. (This is via two lines of descent from Pearl/three great-grandchildren of Mary Caroline, in databases at Ancestry, FTDNA, MyHeritage, and GEDmatch, and with one great-grandchild at 23andMe.) Mary Caroline and Lewis were married just over three months before the birth of their first child. The aforementioned descendants have extremely high/high links to Dockery/Skelton/Porter families in the area who are not descendants of Mary Caroline's theorised sister Talietha Perry who married into those lines. So, in addition to speculation about Mary Caroline's parents, it does not seem unreasonable to stay aware of a possible non-paternal event for Mary Caroline's first child, and possibly one involving her proposed brother-in-law's relatives.
Update: The aforementioned testers now (May 2019) have a dozen unique matches to descendants of William H Porter, husband of Mary Caroline's speculative sister, Talietha, via six children of his first wife. (In this case, "unique" means that all of the matches have different parents if not grandparents.) Seven of the matches are greater than 20 cM, with the three largest (39 cM, 45 cM, 71 cM) being descendants of William's son Aaron Dennis Porter.
Ancestry's "ThruLines" feature also shows tens of matches to the proposed parents and grandparents of William H Porter. The least-distant of these matches are in the 17-23 cM range, but most matches are under 13 cM. However, more research needs to be done to eliminate other connections as well as to verify William's relationship to these suggested ancestors.
Have you taken a DNA test? If so, login to add it. If not, see our friends at Ancestry DNA.