IX Coach

How the Adult Development Stage Models Compare

What each stage map counts, where they have been checked against one another, and where they have not

How do Kegan, Loevinger, Cook-Greuter, Torbert and O’Fallon’s adult development stages line up?

These stage models do not share one scale. The Loevinger lineage of sentence-completion tests counts 8 levels in Loevinger’s original test and 9 in the Cook-Greuter and Torbert family, while O’Fallon’s STAGES counts 12. Kegan’s construct-developmental model is judged by many, Kegan included, to measure a closely related construct, but the authors of the main summary know of no empirical work correlating it with Loevinger’s lineage. The level-by-level crosswalk these sources report, STAGES against Cook-Greuter’s Maturity Assessment Profile (MAP), reports substantial agreement up to level 4.5 in a study whose authors include the model’s developer.

If you have met more than one adult development model, you have probably noticed that the stages do not line up neatly. Loevinger’s sentence-completion tradition, extended by Susanne Cook-Greuter and by William Torbert’s group, and Terri O’Fallon’s STAGES model all describe a path of meaning-making, but they divide it differently and name it differently. Robert Kegan’s subject-object work, a constructive-developmental theory of how adults make meaning, sits nearby and is often placed beside them. This page does not rank these models or say which is correct. It sets out what the main comparative source reports: how many levels each map defines, which correspondences have been measured, which have not, and where the same label is used for different things. Where a correspondence is reported, the page names whose alignment it is.

実践法

これらの実践のつながり

O’Fallon (2020) defines stages as successive developmental perspectives that each transcend and include the previous one, a definition she notes is similar to Kegan’s subject-object research. Where the sources read here compare the models, they differ mainly in where their boundaries fall, how many levels they count, and what the labels are attached to. The practices on this page all come down to keeping those three things visible, so that a number or a name is never read as if it had travelled unchanged from one map to another.

Is Kegan’s stage model the same thing as Loevinger’s ego development stages?

Murray and O’Fallon (2020) report that many researchers, Kegan included, judge Kegan’s subject-object work and Loevinger’s ego development to measure closely related constructs. They also say they know of no empirical work investigating the correlation between the two. So the accurate answer is that they are judged to be closely related, and whether they correspond level by level has not been shown in the sources read here. Placing them side by side is a reasonable starting point for reflection, but it does not establish that a Kegan order equals a Loevinger stage.

How closely does the STAGES model match Cook-Greuter’s MAP, level by level?

In the replication study that Murray and O’Fallon (2020) summarize, STAGES scores were compared with Cook-Greuter/Loevinger scores on the same sentence-completion inventories. The authors report a weighted kappa (a statistic for agreement beyond chance) of 76%, which they place in the substantial-agreement band. Agreement varies by level: a STAGES 3.5 Achiever score matched the MAP level 65.1% of the time and fell within one level 90.7% of the time, and a STAGES 4.5 Strategist score matched 62.2% of the time and within one level 86.5% of the time. The study used four STAGES scorers, one of whom had learned both methods, and its authors include the model’s developer. The correspondence is therefore the authors’ own result, and no independent replication appears in the sources read here.

Are Torbert’s action logics the same as the STAGES levels or the MAP levels?

The sources read here do not show that they are the same. Starr and Torbert (2005) present seven action logics, adapted from Rooke and Torbert’s 2005 work, with percentages drawn from 4,310 respondents; Expert and Achiever are the two largest groups. Murray and O’Fallon (2020) report nine levels for the MAP family, which includes the Global Leadership Profile and Leadership Development Profile instruments associated with Torbert. The sources read here do not reconcile Starr and Torbert’s seven logics with those nine levels, so the counts differ and this page does not claim a mapping between them. The labels Expert, Achiever and Strategist also appear in STAGES, where they name different things on different instruments. The papers read here give no level-by-level crosswalk between Torbert’s logics and STAGES, so any pairing between them is a reading, not a published correspondence.

Why do the level counts differ between Loevinger, Cook-Greuter and O’Fallon?

Each model divides the same developmental territory differently. Murray and O’Fallon (2020) explain that Cook-Greuter extended Loevinger’s system by differentiating two stages, Construct Aware and Unitive, within Loevinger’s final stage. O’Fallon went further and differentiated four stages (5.0, 5.5, 6.0 and 6.5) in the same late territory, and also split the Conformist level into two. That gives 8 levels in the Loevinger test, 9 in the Cook-Greuter and Torbert family, and 12 in STAGES. The extra levels are not all in the later stages. Because the models define levels above 4.5 differently, Murray and O’Fallon say their data did not allow a robust comparison of those levels.

Can a single chart put all of these stage models on one scale?

Not on the evidence in these sources. Each model has its own level count and boundaries. The level-by-level crosswalk these sources report covers STAGES and Cook-Greuter’s MAP up to level 4.5, and Murray and O’Fallon (2020) say that several neighboring developmental models have not been compared thoroughly with one another. A combined chart is an interpretation with its assumptions stated. The honest version marks which pairs were measured against each other and which were placed by reasoning.

どれを選べばよいか

Start from the question, not the chart. If you want to know how a stage name you have heard relates to a number on another scale, read the published crosswalk one row at a time and treat the spread as the answer. If two people seem to disagree about a stage, name each model first and see whether they are using the same word for the same thing. If someone gives you a neat equivalence, ask whether anyone measured it before you adopt it. If a single situation keeps replaying, describe it in two maps and mark where they split. And if a chart arrived without a source, trace it to its paper before you use any of its numbers.

これらのアプローチでは解決できないこと

The one published correspondence in these sources, the STAGES–MAP replication, includes the model’s developer among its authors and used four trained STAGES scorers, one of whom had learned both systems, and the sources read here include no independent replication of it. Murray and O’Fallon (2020) say they know of no empirical work correlating Kegan’s construct-developmental model with the Loevinger lineage, and they note that the neo-Piagetian theories (stage theories built on Piaget’s work) of Commons and of Fischer have not been compared thoroughly with it. Torbert’s seven action logics are read here through a 2005 table; the Harvard Business Review article it adapts was not read, and no level-by-level crosswalk between those logics and STAGES was found in the sources read. Above level 4.5 the models define levels differently, so the sources read here do not support placing the highest stages on one scale. Kegan (1994) was not read for this page, and the quoted definition is the one O’Fallon (2020) reports.

IX Coach でこれを実践する

IX Coach で実践する

よくある質問

How do Kegan, Loevinger, Cook-Greuter, Torbert and O’Fallon’s adult development stages line up?

These stage models do not share one scale. The Loevinger lineage of sentence-completion tests counts 8 levels in Loevinger’s original test and 9 in the Cook-Greuter and Torbert family, while O’Fallon’s STAGES counts 12. Kegan’s construct-developmental model is judged by many, Kegan included, to measure a closely related construct, but the authors of the main summary know of no empirical work correlating it with Loevinger’s lineage. The level-by-level crosswalk these sources report, STAGES against Cook-Greuter’s Maturity Assessment Profile (MAP), reports substantial agreement up to level 4.5 in a study whose authors include the model’s developer.

「Write the model’s name beside every stage you record」はどのように機能するのか

Murray and O’Fallon count 8 levels for Loevinger’s model, 9 for Cook-Greuter’s and Torbert’s models, and 12 for STAGES, so a stage number only means something on the scale it came from. Labels also recur. Starr and Torbert’s seven action logics (in their table, the characteristic ways a leader acts, such as Expert, who “rules by logic and expertise,” and Achiever, who “meets strategic goals”) and STAGES both use Expert, Achiever and Strategist, naming different things on different instruments. Strategist in Starr and Torbert’s table “generates organizational and personal transformation”; in STAGES it is the 4.5 level. A note that carries its model with it stops the label from quietly moving to another scale.

「Pick the map whose resolution matches your question」はどのように機能するのか

A map can only show the distinctions it draws. Murray and O’Fallon report that Cook-Greuter extended Loevinger’s final stage by splitting it into two, Construct Aware and Unitive. Construct Aware is the first MetAware level in O’Fallon’s STAGES (5.0), and the sources read here name Unitive without describing it in detail. O’Fallon’s STAGES splits that late territory into four levels (5.0, 5.5, 6.0 and 6.5) and also splits the Conformist level. On this page’s reading, a question about the late stages is poorly served by a map that gives them one box, and a question about a finer distinction needs the map that makes it. The fit between map and question is a working assumption here, not a tested rule.

「Read a published crosswalk one level at a time」はどのように機能するのか

A crosswalk counts how often two scoring systems place the same inventory at the same level, and the cells show the spread. In Table 1 of Murray and O’Fallon (2020), which reproduces the O’Fallon et al. (2020) replication, a STAGES 3.5 Achiever score matched the Cook-Greuter/Loevinger level 65.1% of the time and fell within one level 90.7% of the time, while a STAGES 1.5 score matched 33.3% of the time. On this page’s reading, reading the row this way shows how far neighboring levels blur, which a ladder diagram tends to hide.

「Ask whether anyone has tested a claimed match」はどのように機能するのか

Murray and O’Fallon (2020) say that Kegan’s construct-developmental work and the Loevinger lineage are judged by many, Kegan included, to measure closely related constructs, and that they know of no empirical work investigating their correlation. On this page’s reading, a claimed equivalence that has never been measured is an assumption, however finished the chart that shows it looks. Asking whether a match was tested separates a measured overlap from a plausible one.

「Keep one shared stage word from carrying two meanings」はどのように機能するのか

Some names recur across these maps. Starr and Torbert’s seven action logics include Expert, Achiever and Strategist, and STAGES also uses Expert at 3.0, Achiever at 3.5 and Strategist at 4.5. On this page’s reading, a conversation can turn on a shared word while the two speakers are pointing at different instruments. Naming the instrument first turns a dispute about who is right into a question about what each person measured.

「Describe one situation in two maps and mark the gap」はどのように機能するのか

Two maps reading the same episode can disagree where their level boundaries fall differently. Writing both readings side by side makes the disagreement something you can look at, instead of a verdict in one voice. On this page’s reading, the place where the readings split points to the distinction the situation depends on. That is a hypothesis to test against the situation, not a result the sources establish.

「Trace a stage chart to the paper behind it」はどのように機能するのか

Many stage charts online combine names, numbers and colors from several models and give no source. The model papers give the counts and names: Murray and O’Fallon (2020) give the level counts for each family, and O’Fallon (2020) gives the STAGES level names and the stage-specific confusions, so a chart can be checked against them one line at a time. On this page’s reading, a chart whose counts or names do not match the originating paper is probably describing something other than the model it names.

Is Kegan’s stage model the same thing as Loevinger’s ego development stages?

Murray and O’Fallon (2020) report that many researchers, Kegan included, judge Kegan’s subject-object work and Loevinger’s ego development to measure closely related constructs. They also say they know of no empirical work investigating the correlation between the two. So the accurate answer is that they are judged to be closely related, and whether they correspond level by level has not been shown in the sources read here. Placing them side by side is a reasonable starting point for reflection, but it does not establish that a Kegan order equals a Loevinger stage.

How closely does the STAGES model match Cook-Greuter’s MAP, level by level?

In the replication study that Murray and O’Fallon (2020) summarize, STAGES scores were compared with Cook-Greuter/Loevinger scores on the same sentence-completion inventories. The authors report a weighted kappa (a statistic for agreement beyond chance) of 76%, which they place in the substantial-agreement band. Agreement varies by level: a STAGES 3.5 Achiever score matched the MAP level 65.1% of the time and fell within one level 90.7% of the time, and a STAGES 4.5 Strategist score matched 62.2% of the time and within one level 86.5% of the time. The study used four STAGES scorers, one of whom had learned both methods, and its authors include the model’s developer. The correspondence is therefore the authors’ own result, and no independent replication appears in the sources read here.

Are Torbert’s action logics the same as the STAGES levels or the MAP levels?

The sources read here do not show that they are the same. Starr and Torbert (2005) present seven action logics, adapted from Rooke and Torbert’s 2005 work, with percentages drawn from 4,310 respondents; Expert and Achiever are the two largest groups. Murray and O’Fallon (2020) report nine levels for the MAP family, which includes the Global Leadership Profile and Leadership Development Profile instruments associated with Torbert. The sources read here do not reconcile Starr and Torbert’s seven logics with those nine levels, so the counts differ and this page does not claim a mapping between them. The labels Expert, Achiever and Strategist also appear in STAGES, where they name different things on different instruments. The papers read here give no level-by-level crosswalk between Torbert’s logics and STAGES, so any pairing between them is a reading, not a published correspondence.

Why do the level counts differ between Loevinger, Cook-Greuter and O’Fallon?

Each model divides the same developmental territory differently. Murray and O’Fallon (2020) explain that Cook-Greuter extended Loevinger’s system by differentiating two stages, Construct Aware and Unitive, within Loevinger’s final stage. O’Fallon went further and differentiated four stages (5.0, 5.5, 6.0 and 6.5) in the same late territory, and also split the Conformist level into two. That gives 8 levels in the Loevinger test, 9 in the Cook-Greuter and Torbert family, and 12 in STAGES. The extra levels are not all in the later stages. Because the models define levels above 4.5 differently, Murray and O’Fallon say their data did not allow a robust comparison of those levels.

Can a single chart put all of these stage models on one scale?

Not on the evidence in these sources. Each model has its own level count and boundaries. The level-by-level crosswalk these sources report covers STAGES and Cook-Greuter’s MAP up to level 4.5, and Murray and O’Fallon (2020) say that several neighboring developmental models have not been compared thoroughly with one another. A combined chart is an interpretation with its assumptions stated. The honest version marks which pairs were measured against each other and which were placed by reasoning.

研究

関連する概念