Horizontal and vertical family relationships, literarily

Aunt and uncle yield to grandmother and grandfather.

In my Sociology of Family class, I ask the students to raise their hands if they experienced a substantial direct relationship with a great-grandparent. As all eight of my great-grandparents predeceased me, I am surprised that so many student raise their hands. (Maybe elite college students come from families with better health and healthcare, but still.)

Demographically speaking, the decline of fertility and the increase in survival to old ages has meant that we have fewer horizontal (intragenerational) relationships and more vertical (intergenerational) relationships. And these intergenerational relationships are also more intensive, lasting years and decades.

And so, for the question of the day: Does this shift from intragenerational to intergenerational show up in the giant Google book word database? As a one-day expert in ngramology, I can conclude: Mostly, yes.

On the closest relationships, brother, sister, mother, father, the evidence is pretty good. Since 1800, occurrences of “mother” have increased from about three-times to more than four-times occurrences of “sister.”

On “father” versus “brother,” both show declines, but the ratio increased from about 1.5-to-1 to 2.5-to-1.

The picture for aunt/grandmother and uncle/grandfather is more clear. “Grandfather,” representing intergenerational, has almost completely closed the gap on “uncle,” which is one degree more horizontal:

Even more strongly, “grandmother” has vaulted past “aunt”:

You can experiment with variations here.

For an interpretation of demographic trends and aging for family relationships, consider this issue of  Daedalus and a variety of recent books, some taking global perspectives, some focusing on intergenerational support challenges.

10 thoughts on “Horizontal and vertical family relationships, literarily

  1. Neat insights, Phil. I was really intrigued by the social implications of the uncle/aunt versus grandparents shift, and pondering what kinds of differences one experiences in terms of mentoring and loyalty as a result. Do we learn different kinds of civic behavior or negotiation strategies as a consequence? Does it affect our creativity and risk-taking behavior?

    Also, what about step-relations?


  2. I can’t evaluate whether or not your statements about vertical versus horizontal relationships is true, but for the sake of argument, I’ll grant that it is. Given that, I don’t find it convincing at all that the n-gram results are evidence for this trend.

    First, why should these changes be reflected in the google books corpus? If all the books were personal diaries, then the case would be stronger, but they’re not. There is just no strong linking hypothesis between these structural changes in families, and the frequency with which these family words should appear in a corpus of randomly selected printed material. A better control would be to compare “my brother” to “my father.” These results would be much clearer cases of people talking about real relationships.
    The case would also be stronger if you could demonstrate that the change in ratios was chronologically related to the kind of change you are discussing. As it stands, you’re just comparing the two most different time points to each other, which are completely arbitrary and may not be significant time points for these structural changes you are discussing.

    Secondly, if you could assume some reasonable linking hypothesis between these structural changes and the words people use, it’s not clear that the data you presented reflect these changes. Take the comparison between aunt~grandmother and uncle~grandfather. Both of these are vertical relationships, and the grandparents are one extra degree vertical. If it is the case that relationships are extending up and up, and that people use more words describing these relationships, we should expect to see that the frequency of grandparent words increase relative to aunt and uncle. However, the strong trends in the charts you show is the decrease of aunt and uncle words. How does this follow from the hypothesis that increasingly vertical familial relationships are reflected in frequency of words for those relationships? If not, how do you account for it without resorting to story telling?

    Finally, it’s not clear that the trends you’ve presented are really real. The n-gram corpus is case sensitive, so you should really compare aunt+Aunt ~ grandmother.


    1. Josef – Thanks for your comments. Coupla points.

      First, we already know there is a trend from horizontal to vertical relationships, which follows almost automatically from the declines in fertility and increases in life expectancy (fewer sibling/aunts/uncles/cousins, more grandparents/greatgrandparents). I say “almost” because there is a possible countertrend from the growth of step/blended families, which add horizonticality to the family tree. Anyway. So you may have missed the direction of my inference. The question was, “Does this shift from intragenerational to intergenerational show up in the giant Google book word database?,” not, “Does the book database confirm the existence of this trend?”

      On vertical/horizontal: aunt and uncle are not just less vertical than grandmother/grandfather — they are also more horizontal, because they refer to sibling relationships. Everyone has two biological parents and four grandparents who may be living, but if everyone only has one child there are no aunts and uncles.

      Second, your methodological points are well taken. Anyone studying this seriously (rather than writing a blog post several days after the database’s debut) should take all that into account. The suggestion about “my aunt,” versus “aunt” is one interesting way of getting at different uses. Just using “aunt” includes all kinds of literary as well as non-fiction references. That is a simple thing to look at, and it appears that the patterns are generally the same, with “my grandmother” and “my grandfather” surpassing “my aunt” and “my uncle”: http://bit.ly/eMDqlv.


      1. I wasn’t trying challenge your statements about family structure. They’re completely outside of my expertise, so I’ll just accept them as true without any further need for empirical support. I guess I worded my comments as “the n-gram data isn’t evidence for this,” but I really did mean to direct them at your intended inference.

        For instance, families are more mobile now than they used to be (at least, that’s my pop-sociology understanding). I would expect the familial relationship to take the biggest hit due to this mobility to be aunts, uncles and cousins. Holding fertility constant even, you might still expect to see the drop in aunt and uncle due to mobility. Could mobility alone cause a drop of the magnitude observed in the data? Does mobility plus fertility have to be appealed to in order to account for the magnitude of change? Neither of us can answer these questions, because there no prior linking hypothesis about how we should see societal changes like these reflected in the data. Not even a sketchy one.

        And that’s the problem when we’re just trying to arbitrate between the two stories you and I have been clever enough to come up with. Perhaps what’s really going on is a growing social taboo that we’re not consciously aware of about talking about aunts and uncles, or many other things.

        That’s not to say I don’t think there are interesting patterns in the data. For example, looking at “my aunt”, “my uncle” and “my cousin”, something clearly happens between 1900 and 1920, with a steep decline in all, followed by a brief blip between 1920 and 1930-ish, then another gradual decline. I’d like to know what caused that, but I think that all possible hypotheses are equally likely until the trends are strongly linked to specific historical events, or other quantitative trends.

        As for the fertility-longevity hypothesis, there isn’t exactly convergent evidence. Brother~uncle, sister~aunt show very different profiles. You might expect a dip in brother preceding a dip in uncle by a few decades, but you don’t. Also, you would expect to see decreasing uses of things like “my brothers”, “my sisters”, “my aunts” and “my uncles.” Looking at “my brothers” and “my sisters”, the trend looks basically flat: http://bit.ly/fjk1Zg and looking at “my aunts” and “my uncles,” the trend is actually positive: http://bit.ly/dPh8Ws

        What I’m trying to say is that I believe you that the trends in families are real, but there is no a priori reason to assume they should be reflected in the use of familial words in the n-gram corpus, and based on the evidence you’ve shown here, and the evidence I’ve looked at, I don’t believe they are.


  3. My guess is that the word count does not exactly “reflect” the reality in the census. Yes, lower fertility means that people have fewer aunts and uncles, and longevity means that more people have living grandparents. But I don’t think the graphs of relatives in books track so closely with graphs of living relatives.

    Relatives in books are relatives that people think are important enough to write about. Maybe the decline in aunts and uncles reflects not so much a change in their numbers but a change in their importance in people’s lives. Of course, “importance in people’s lives” is hard to know. But we could use “presence in people’s homes” as a proxy. If we graphed census data on uncles and aunts living in the household (I think the census has that info, doesn’t it?), would it parallel their decline in books?


    1. That’s my thinking, too, Jay. It’s hard to be sure with older Census data, but the overall trend is from horizontal toward vertical (and from extended to nuclear, though since 1980 that’s rebounded some).


Comments welcome (may be moderated)

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s