* What is an objective definition of "Splitting"?

[Gary_G]

Before I get too far into my data re-entry as a "Splitter", I want to make sure I have an objective definition of "Splitter" and stick to it.

With templating one has the ability to store data in the source and citation portions. That seems to have blurred the definition of what appeared to be two distinct approaches; "Splitting" and "Lumping". So; I thought I'd into it a bit more and post my findings for feedback.

I had the feeling that an objective definition of "Splitting" was likely based on the GEDCOM standard format. So I took a look at an old copy of the draft 5.5.1 standard that I had on hand...

! Added a <<SOURCE_CITATION>> structure subordinate to the fact being cited. It is generally best if the source citation contains only information specific to the fact being cited and then points to the more general description of the source, defined in a SOURCE_RECORD. This reduces redundancy, provides a way of controlling the GEDCOM record size, and more closely represents the normalized data model.

! Systems that describe sources using the AUTHor, TITLe, PUBLication, and REPOsitory fields can and should always pass this information in GEDCOM using a SOURce record pointed to by the <<SOURCE_CITATION>>. Systems that only allow free form source notes should encourage forming the source information so that it include text about these categories:

! TITL: A descriptive title of the source
! AUTH: Who created the work
! PUBL: When and where was it created
! REPO: Where can it be obtained or viewed

When possible provide the tag for these categories within the text so that a receiving system could parse them to fit the recommended source/citation structure..

It seemed to imply that the original GEDCOM intent was to have a citation-specific portion to each source. That is; for a book source, the citation-specific portion would refer to the pages of the book. I've seen this structure expressed in "freeform" templates, which I understand is the closest that a program can come to producing GEDCOM-standard records.

If this is the case, then it would appear that "splitting", in its purest sense, would document a page in the citation-portion and the containing document in the source portion. That would imply that "Splitters" would provide the page reference, attach the page image and enter the page text into the citation-portion.

This is quite different from what I'd understood from reading various articles that talk about "Splitting". They state that "Splitters" put absolutely everything in the source-portion and not even enter a citation-portion.

So; if I'm going to do "Splitting", I'm now wondering if I should revise the way I enter my information to follow the definition implied by the GEDCOM Standard...

Are there issues peculiar to the operation of FH7 that would impact adopting a more GEDCOM-based approach?

[ColeValleyGirl]

If this is the case, then it would appear that "splitting", in its purest sense, would document a page in the citation-portion and the containing document in the source portion. That would imply that "Splitters" would provide the page reference, attach the page image and enter the page text into the citation-portion.

This is quite different from what I'd understood from reading various articles that talk about "Splitting". They state that "Splitters" put absolutely everything in the source-portion and not even enter a citation-portion.

Common sense has to come into it and Splitting/Lumping is a continuum, not two discrete poles.

Take a multi-volume diary. I would define each volume as a source, and use the citation portion to specify:

• where in the volume the relevant information is found (specified by date or page, depending on how the diary is structured).
• a transcription or abstract of the relevant info
• a quality assessment
• maybe a note e.g. to say why this particular piece of info is relevant to the assertion I'm citing it for

So you could describe that as 'lumping at the volume level' or 'splitting at the volume level'. Somebody who prefers to lump as much as possible might define the whole diary as the source, and record the volume as well as the date/page in 'where within source'. Somebody who prefers to split as much as possible might define a single diary entry as the source... but if it was necessary to cite multiple diary entries they might adopt the 'volume' or 'diary-based approach on purely practical grounds.

Moreover, as the citation is the only place you can assign a Quality Assessment, those splitters for whom that's important will never have a totally 'empty' citation.

None of this is new with templates. And I doubt you'll find a 'Splitter' who doesn't have some degree of lumping for some sources.

[jbtapscott]

And I doubt you'll find a 'Splitter' who doesn't have some degree of lumping for some sources.

I think this hits the nail on the head!!. I just wonder sometimes if these "principals" are being over thought. There will be occasions where I have a book as a Source and then citations in multiple facts pointing to that source, with each citation having a chapter number (if relevant) and a page number, etc., (similar to what Helen describes). I also have sources (e.g. Census) where there is nothing in the multiple citations pointing to the source record.

I don't class myself as a "splitter" or a "lumper" (or whatever description is being applied this month!!) but as somebody who wants to record data / facts, etc accurately and consistently, with the minimum amount of repetition as possible.

[LornaCraig]

I agree with Helen, splitting/lumping is a continuum. The level of granularity at which you choose to ‘split’ is flexible.

One practical consideration is the linking of images. In early versions of FH it was not so easy to link media items to citations and when they were linked to citations they were not very ‘visible’. For this reason some users who have been with FH from early days tend to use split sources whenever they have a media to link, but lumped sources for cases where they don’t need to link media – for example BMD indexes.

The distinction is still reflected in the FH Essentials collection of templates, where most of the descriptions say something like “One source record per register entry/certificate/census household…” but for the Civil Registration Index it says “One source per index”.

FH has adapted, partly in response to users who have imported from other products where ‘lumping’ was the norm, so it is now easier to link media to citations. But a new user who takes FH ‘out of the box’ will probably find themselves using the Essentials templates as they are designed, without spending a lot of time agonising over ‘to split or not to split’. This is an example of the kind of discussion new users (not importing from other programs) probably find confusing.

[Gary_G]

Helen;

Thank you. I understand what you are saying and actually noted the blurred lines between lumping and splitting in my post.

I suppose that what was bothering me is that I had been told "Splitters" do not to put anything in the citation portion of a template. From what you've said, this is incorrect. The GEDCOM standard also seems to imply that that is not the case. That is; the GEDCOM standard intended for there to always be a citation-portion that at least contains the "Where-in" portion of a citation.

For me; this casts a very different light on the definition of a "Splitter"; at least in it's role as one end-member of the spectrum.

I tend to divide up my citations based upon how they are stored. If a census is referenced as a stand-alone entity on a website, then I reference the whole census as a source and the census page as a citation. If the census is just an element of a referenced film, then I use the film as a source and the specific frame as the citation. Essentially; I tend to "back up" one level in the reference structure from the "where-in" and call that the source. Seems to me that this results in consistent use of the citation-portion and implies that I should typically attach the image I am citing to the citation-portion and also use the citation notes for any transcription.

I currently have a working approach in which a repository contains the invariant information on a website and dataset/collection. That information is referenced by a set of customized generic templates. So; I usually get a fair amount of reusability, while generating compliant Evidence Explained citations and leveraging all the FH7 benefits of a splitter approach. Defining "Splitting" in a bit more rigorous way is just the last piece of the puzzle. It should allow me to write down exactly how I plan to enter my data, so I don't go wandering off at some point.

By the way; I finally found a post on, "Understanding Source Templates" https://fhug.org.uk/forum/viewtopic.php?f=32&t=21211
It was rather interesting, as it comes at the topic from a totally different direction.
[Almost didn't find it, due to the way a set of search terms is processed by the board. ]

[Jean001]

I suppose that what was bothering me is that I had been told "Splitters" do not to put anything in the citation portion of a template.

As with so many of these discussions there is no 'right' or 'wrong' way. Some 'splitters' add nothing to a Source Citation, others add something.

[Gary_G]

Lorna;

While the overall topic is possibly confusing, some of what you stated is exactly what new users need to hear in order to not set themselves on a path that they may later want to reconsider. As you noted; the typical usage of FH has a lot of history behind it and some things changed significantly with the advent of FH7. This is something that a new user, relying on older posts, might not notice. Also; long-time users who do something a particular way due to historical FH7 constraints, often don't mention to "newbies" that they now have the option to consider another path.

Just to make sure I understood you... If one attaches images and transcriptions to the citation-portion in FH7, should they now be just as visible as if they were attached to the source-portion? [This includes making them visible in reports by setting the option-switches of the report.]

[Gary_G]

Jean;

I was VERY careful to not even imply that one way was "right" or "wrong".
I am simply interested in how the term might be objectively defined; possibly based on what the GEDCOM standard supports.
This was a logical approach, as much of what FH7 does is based on that standard.
Being clear on that particular definition helps me decide what is the data-entry approach I wish to take.

[Gary_G]

"jbtapscott";

From the start, I essentially recognized that Splitting<->Lumping is a spectrum. But that was not the focus of the post.
I am simply trying to figure out what is best for me.
For me to make an informed decision, I needed to find an objective way of defining the "Splitter" end-member of the spectrum.

[AdrianBruce]

Like Helen, while I class myself as a splitter, there are many cases where I record a lumped source simply because the extra effort to split it, doesn't seem to add anything - a City Directory, for instance. A source record per entry? No thanks!

With that in mind, I think that my attempt at a definition of split sources is something like this:

The source record should describe that part of a physical object (containing useful genealogical information) where the part is as small as possible, without losing information derived from the arrangement of the data in that or adjacent parts.

Caveat 1 - the "physical object" is not the screen of my PC, nor is it even a digital image on a server somewhere. Rather it's the physical object that was used to create the JPG on my screen.

"where the part is as small..." To illustrate this, if I were using a baptismal parish register - or probably the images of it - I can split what the source record represents until I get to the smallest chunk describing a single entry, because each entry is independent (Caveat 2 coming up...) That single entry forms my source record. I can't slice the source record up further because if I did, and I had a source record for each "cell" of data, I'd be losing the information of how the individual words stack together. E.g. what use is the name of a child without the date of baptism?

Caveat 2 - if I have a situation where a family "saved up" their children to be baptised all in one job lot, then, although that's (certainly in later years) several entries in the register, I would record all that in one source record because it's an essential part to realise that the parents went along (or were made to go along!) with all their children in one go - this information is more difficult to convey if it's several source records.

One essential part of my workflow is to record why I think this source applies to this person ("proof of identity"). That will normally be recorded as part of the Notes in a Source Record. If I try to put it into the Note of the Citation-Specific Detail, it's (a) hidden a layer below and (b) there's less room to see it - at least in my set-up. On that basis, I find splitting naturally allows the "proof of identity" to go into the Source Record. (In reality, this is probably one good reason why I went for splitting).

As I said though, there are plenty of occasions where I use a lumped source. Indexes of vital events, tend to end up as a lumped source. Wikipedia is one such example - I use it for regimental histories, etc, and just have one source record for Wikipedia. The effort to multiply source records just seems like too much hard work - all the relevant data goes into the Citation-Specific Detail. Ditto a City Directory. FreeBMD (an index of civil registration events in England & Wales) is also a single Source Record to save me from typing in its URL yet again. However, I am inconsistent on this - I seem to create Source Records for individual entries in the California Death Index, for instance. I think this is because I seem to end up with more text in the "proof of identity" and therefore prefer to make it visible in a separate Source Record.

What I think this implies is that my definition of split sources is about my workflow, it's not data-based. And it's not consistent on my part, as the California Death Index (usually a source record per event) compared to FreeBMD (one single source record) shows.

[LornaCraig]

I needed to find an objective way of defining the "Splitter" end-member of the spectrum.

For me, I always go to the 'splitter' end of the spectrum when:

(a) I have media to link. This is partly for historical reasons - as I said in my previous post, in early version of FH it was much easier to 'see' media at Source level.

(b) putting the transcript of the relevant text in the Source record avoids duplication of the text in every citation. And if the transcript needs changing it only has to be done once.

I tend to divide up my citations based upon how they are stored. If a census is referenced as a stand-alone entity on a website, then I reference the whole census as a source and the census page as a citation. If the census is just an element of a referenced film, then I use the film as a source and the specific frame as the citation.

For me censuses are at the 'splitter' end of the spectrum. I 'split' down to the household level. I have always used AS for recording census data and from the start it encouraged using 'Method 1' (split) sources for censuses. But even if I hadn't used AS I would still be a full splitter for censuses, for the above reasons.

Just to make sure I understood you... If one attaches images and transcriptions to the citation-portion in FH7, should they now be just as visible as if they were attached to the source-portion? [This includes making them visible in reports by setting the option-switches of the report.]

Yes, I think so (but as I don't do things that way I can't be absolutley sure). But I would still want to avoid duplication of transcripts in multiple citations. I know some people object to duplication as a matter of principle. For me it's a personal preference.

[Gowermick]

I am simply trying to figure out what is best for me.
For me to make an informed decision, I needed to find an objective way of defining the "Splitter" end-member of the spectrum.

When I started using FH, I used to quote a Parish Register image for a Christening as "St George In the East Register Image" or "Limehouse St Ann Register Image" and stored the image to my PC with a title like "J0036 Christening of John Smith". As a lumper, I used the source "St George In the East Register Image" as many times as necessary, with nothing added in the citation at all.

As the number of Church registers grew, I found I was wasting more and more time finding the source for an existing church name, or if it wasn't there, creating a new one for the church. Then the penny dropped, that because I was already storing church name within the fact, I didn't need to repeat it in the source title, especially as it didn't help finding the image (their naming didn't include the church name).

So then I adopted just three basic sources "Christening Image", "Marriage Image" and "Burial Image" and use these for every Christening, Marriage or Burial whenever I found an image. ( Now That's what I call true lumping )

This is the reason I became a Lumper i.e to save me time and effort creating sources.

I learnt long ago that quoting exactly where I found the image was a complete waste of time, because a) I had the image, so didn't really need to know where I got it from, and b) websites are continually being adjusted, so what is true one day, may be an "Error 404" the next.

[Gary_G]

Adrian;

Thank you for your suggested definition.

I recognize that practicalities may cause one to move along the Splitter<->Lumper spectrum for specific records. However; a spectrum only has meaning when its end-members are well defined. Based upon what I've seen so far, the GEDCOM standard expects to have at least some content in the citation portion. So; even the most extreme splitter ought to have something in the citation-portion. That is the aspect that was bothering me the most, since I'd been told that the most extreme splitting would result in having absolutely no content in the citation-portion. Based on what I've been told in this thread, that viewpoint was predicated on the capabilities of a precursor to FH7 and is no longer valid.

Sidenote...
The above brings up a significant point about the danger of not identifying a post with an FH version to which the post pertains. As a new-user (even though familiar with other programs) I was unable to weed out outdated comments in old posts on the forum. This resulted in a mistaken impression of how the current FH version, FH7, should/could be used. While some older posts are generic and still applicable, I can see quite a few people falling into the same "trap" of following the advice in an outdated post.

[NickWalker]

Based upon what I've seen so far, the GEDCOM standard expects to have at least some content in the citation portion. So; even the most extreme splitter ought to have something in the citation-portion. That is the aspect that was bothering me the most, since I'd been told that the most extreme splitting would result in having absolutely no content in the citation-portion. Based on what I've been told in this thread, that viewpoint was predicated on the capabilities of a precursor to FH7 and is no longer valid.

This isn't the case. Some users can choose to record data in the citation when using a 'splitter' source, but there isn't any requirement to do so. I don't have any data that I need to record in citations and even if I did I don't like having duplication of data so would avoid it. But as I keep on stressing in every post and video I've made on this, it is the choice of each user of FH to decide how they want to record their data.

I don't use Templated sources, but if I did I'd use the built in Family Historian 'Essential' source templates for the key sources such as Census and parish records. And they don't have any citation fields.

[Gary_G]

Lorna;

I read over your most recent post. It left me slightly confused and I "think" I know why. Are you are looking at a citation on an entry-by-entry basis?

Much as I hate quoting Evidence Explained, to me, it makes one reasonable point about the subject of citation; "we cite what we see". So I tend to follow that paradigm. Even in the most extreme splitting, I would have "seen" the whole page, or even more than one page (for families spread across pages). That would mean that I would place all that I "saw" in the citation-portion. That is; I would attach the relevant page-images and transcribe them in the citation-portion. This would not result in multiple records to change, since the one citation would still be attached to each affected person's "fact".

However; this approach does still have its issues. Because of the way FH7 works, there is a potential issue with documenting the date one "saw" the pages. That info can't be in the citation-portion, if it is to be used in the bibliography. And yet; one may not have "seen" all the pages on the same day. This presents some irritating issues that require work-arounds to produce fully-compliant Evidence Explained citations. So; I think Calico Pie has some issues yet to be solved. [I know this has been mentioned in a wish-list item.]

[Gary_G]

Nick;

From the same GEDCOM standard document previously quoted:

Added a <<SOURCE_CITATION>> structure subordinate to the fact being cited. It is generally best if the source citation contains only information specific to the fact being cited and then points to the more general description of the source, defined in a SOURCE_RECORD. This reduces redundancy, provides a way of controlling the GEDCOM record size, and more closely represents the normalized data model.

This fairly specific statement in the GEDCOM standard seems to contradict your assertion:

This isn't the case. Some users can choose to record data in the citation when using a 'splitter' source, but there isn't any requirement to do so. I don't have any data that I need to record in citations and even if I did I don't like having duplication of data so would avoid it. But as I keep on stressing in every post and video I've made on this, it is the choice of each user of FH to decide how they want to record their data.

Of course; one can always debate the issue of whether a user "can" or "should" choose to do something...

[NickWalker]

Nick;

From the same GEDCOM standard as previously quoted:

Added a <<SOURCE_CITATION>> structure subordinate to the fact being cited. It is generally best if the source citation contains only information specific to the fact being cited and then points to the more general description of the source, defined in a SOURCE_RECORD. This reduces redundancy, provides a way of controlling the GEDCOM record size, and more closely represents the normalized data model.

This fairly specific statement in the GEDCOM standard seems to contradict your assertion:

This isn't the case. Some users can choose to record data in the citation when using a 'splitter' source, but there isn't any requirement to do so. I don't have any data that I need to record in citations and even if I did I don't like having duplication of data so would avoid it. But as I keep on stressing in every post and video I've made on this, it is the choice of each user of FH to decide how they want to record their data.

Of course; one can always debate the issue of whether a user "can" or "should" choose to do something...

No it doesn't contradict it. You are misunderstanding. It is correct in saying that the data in the citation should be limited to avoid duplication and that's what i do - I limit it by recording nothing in the citation - this is the ultimate way to avoid duplication. The definition you quoted didn't say that you HAVE TO record additional data in a citation.

[ColeValleyGirl]

Lorna;
Much as I hate quoting Evidence Explained, to me, it makes one reasonable point about the subject of citation; "we cite what we see".

Yes, but that's in the context of not citing an original parish register (say) when what we've actually seen is a derivative (be it image, or transcription). It's nothing to do with citing the entirety of what we see.

[LornaCraig]

Lorna;

I read over your most recent post. It left me slightly confused and I "think" I know why. Are you are looking at a citation on an entry-by-entry basis?
... I would attach the relevant page-images and transcribe them in the citation-portion. This would not result in multiple records to change, since the one citation would still be attached to each affected person's "fact".

No, it doesn't result in multiple records to change (if you need to correct a transcripton error, for example). But it results in multiple citations to change. FH doesn't have the concept of a 'shared' citation. In FH citations are all separate, even if they happen to be identical. In a census record there may be several people in a household. If you put the transcription of the household in the citation then you have multiple copies of the transcription because there will be at least one citation for each person, and probably more if you cite the same source for occupation, age, etc, as well as the census itself. So the actual transcription of the text will be stored multiple times in your gedcom. I suppose you could just put each individual's single line in their transcription but that would be unusual.

[Gary_G]

Nick;

I disagree with your interpretation of the standard and am not misunderstanding what I read.
The phrase "generally best" has a very important meaning when used in a standard and would not have otherwise been included.
Such a phrase is sometimes used when an explicit "should" or "shall" clause is unenforceable, which is definitely the case with a standard such as GEDCOM.
It is also used when it is felt that deviating from what is written may lead to unintended results.
This precisely why some genealogy programs on the market generate GEDCOMs which are not easily transferrable.
Those developers have often taken liberties and loosely interpreted what was written.

Of course; no one HAS TO follow what is actually written in the standard.
Some prefer to weather the consequences.

[Gary_G]

Helen;
Much as I hate quoting Evidence Explained, to me, it makes one reasonable point about the subject of citation; "we cite what we see".

Yes, but that's in the context of not citing an original parish register (say) when what we've actually seen is a derivative (be it image, or transcription). It's nothing to do with citing the entirety of what we see.

Lorna;
I wouldn't have stated it as I did had I not been raked over the coals by the author for interpreting it in a different manner.
I've no wish to go through that again.

[LornaCraig]

Lorna;
Much as I hate quoting Evidence Explained, to me, it makes one reasonable point about the subject of citation; "we cite what we see".

Yes, but that's in the context of not citing an original parish register (say) when what we've actually seen is a derivative (be it image, or transcription). It's nothing to do with citing the entirety of what we see.

Lorna;
I wouldn't have stated it as I did had I not been raked over the coals by the author for interpreting it in a different manner.
I've no wish to go through that again.

Please note that it was Helen (ColeValleyGirl) not me, to whom that reply should be directed. She knows far more about Evidence Explained that I ever will!

[Gary_G]

My sincere apologies, Lorna. These topics get rather confusing after a while with the multiple responders chiming in.
I tried to amend the post when I noticed the error, but the forum wouldn't permit it.

[Gary_G]

Lorna;

I read over your most recent post. It left me slightly confused and I "think" I know why. Are you are looking at a citation on an entry-by-entry basis?
... I would attach the relevant page-images and transcribe them in the citation-portion. This would not result in multiple records to change, since the one citation would still be attached to each affected person's "fact".

No, it doesn't result in multiple records to change (if you need to correct a transcripton error, for example). But it results in multiple citations to change. FH doesn't have the concept of a 'shared' citation. In FH citations are all separate, even if they happen to be identical. In a census record there may be several people in a household. If you put the transcription of the household in the citation then you have multiple copies of the transcription because there will be at least one citation for each person, and probably more if you cite the same source for occupation, age, etc, as well as the census itself. So the actual transcription of the text will be stored multiple times in your gedcom. I suppose you could just put each individual's single line in their transcription but that would be unusual.

Lorna;
FH7 has such different terminology from what I'm used to seeing outside the forum.
The program documentation uses "citation" to refer to something associated with a "source" (ie. a GEDCOM type of definition)
The rest of the world uses a "citation" as essentially equivalent to a footnote.
"Record" seems to be similarly ambiguous due to what I've seen used elsewhere.

I sense that your point is rather important and would really like to understand it fully.
Would you mind briefly describing what you mean by "record" and "citation"?

[NickWalker]

Nick;

I disagree with your interpretation of the standard and am not misunderstanding what I read.
The phrase "generally best" has a very important meaning when used in a standard and would not have otherwise been included.
Such a phrase is sometimes used when an explicit "should" or "shall" clause is unenforceable, which is definitely the case with a standard such as GEDCOM.
It is also used when it is felt that deviating from what is written may lead to unintended results.
This precisely why some genealogy programs on the market generate GEDCOMs which are not easily transferrable.
Those developers have often taken liberties and loosely interpreted what was written.

Of course; no one HAS TO follow what is actually written in the standard.
Some prefer to weather the consequences.

What you are saying defies logic. You seem to be suggesting that we have to record some data in a citation - I have nothing to record in there so why on earth would a standard say I have to? The sentence you have quoted means, if there is information to record in the source citation then "It is generally best if the source citation contains only information specific to the fact being cited". You have definitely misinterpreted this. Here's the standard:

Source Citation.png (70.57 KiB) Viewed 1501 times

As that clearly shows - the only thing that is required (1:1) is the link to the source. All the fields are 0:M or 0:1 which means they are all optional. Which of those fields are you telling me I have to record?! There is nothing in the standard that says that and certainly not that sentence you have misinterpreted.