What Does Thomas PPA Self Image Graph III Describe?

What does “Self Image” name in Graph III?

In a Thomas Personal Profile Analysis (PPA) report, “Self Image” is the provider’s label for Graph III, one of three separately named graph views. It tells you which report field you are looking at. By itself, the phrase does not establish a broad psychological measure of self-esteem, a rating supplied by another person, or a complete account of someone’s self-concept. The ordinary words invite those readings, but the available instrument documentation supports the narrower, report-specific one.

The official Thomas API reference, “PPAAssessment Members, Thomas API Webservices Reference Guide,” names Graph I “Work Mask,” Graph II “Behaviour Under Pressure,” and Graph III “Self Image.” It lists the three as distinct fields. That separation establishes that the report presents differently labeled views; the field names do not explain why their displayed patterns might differ, or prove a cause for any difference. The same reference separately lists “Self-Image Pattern” as another field and gives CSI as an example, which is a reminder that a phrase resembling Graph III’s label can name a different report element. Graph III’s label should therefore be read in its exact location, rather than treated as a general definition of self-image.

The provider’s “PPA Profile Sample Report 2018” makes that layout concrete. Its Graph III section is headed “Self Image,” displays D, I, S, and C, and is followed by interpretive prose. Graph I and Graph II appear as their own sections elsewhere in the sample. The narrative shows how the provider presents the graph in a worked report; it does not turn the heading into an independent measure or establish that a real person has been observed in the ways the prose describes. The PDF is illustrative provider material, not a documented case study.

So, if you encounter “Self Image Graph III,” the defensible first reading is simply: this is Thomas’s named Graph III view in the PPA report. Do not infer from the title alone that it measures self-worth, confirms how others see the person, or settles how the person behaves across settings. The label identifies a place and perspective within the report. What the plotted response format contains, and what conclusions its numbers can support, are separate questions.

Sources: PPAAssessment Members, Thomas API Webservices Reference Guide; PPA Profile Sample Report 2018

How do the PPA responses become a four-scale profile?

The Thomas PPA profile represents a constrained response task: according to the provider’s 2024 technical summary, “The science behind the Thomas Behaviour assessment,” the questionnaire has 24 items, each presenting four adjectives. For each set, the respondent selects the adjective most like them and the one least like them. Thomas describes the resulting profile through Dominance, Influence, Steadiness, and Compliance, with brief work-related descriptions organized around power, people, pace, and policy.

This is a forced-choice format: the person chooses among options supplied together rather than assigning an independent rating to every adjective. The distinction matters for reading the task. A choice says which option was selected as relatively most and least like the respondent within that particular set. It does not record an unrestricted judgment such as how strongly each word fits, and it is not a direct sample of behavior observed during work. The provider calls the PPA a self-report; its documented input is a set of selections made by the respondent.

For a simple hypothetical illustration—not an actual PPA item—imagine four adjectives on a page and a direction to choose one that feels most like you and one least like you. A respondent must make both choices even if two options seem similarly fitting or none feels especially apt. This illustrates the constraint in the response task; it says nothing about which adjectives the real assessment uses, what a particular person would select, or how those selections are transformed into graph values.

Thomas’s four scale names organize the profile into broad work frames. In the 2024 summary, Dominance is associated with power, Influence with people, Steadiness with pace, and Compliance with policy. Those descriptions tell the reader the provider’s intended organizing vocabulary: four named dimensions presented as relevant to work behavior. They do not, on their own, define each dimension exhaustively or establish that every work situation can be reduced to one of these frames. Nor does a scale name make a value a measure of ability, performance, preference in every context, or personal worth.

Because Graph III is built from selections among alternatives, it is reasonable to infer that Thomas applies some structured scoring procedure to produce a four-scale representation. The response format and the existence of four plotted scales make that much clear at a high level. But the public 2024 summary does not supply enough detail to reproduce the scoring algorithm or to interpret each displayed value as a particular quantity. It does not tell a reader, from the task description alone, what units a graph uses, what change in a value means, or how close two plotted values need to be before they should count as meaningfully different.

The same boundary applies to comparison. A forced-choice questionnaire cannot be declared ipsative or norm-referenced merely because respondents choose between options; those are claims about how scores are constructed and interpreted, and require documentation beyond the item format. The available description does not identify a comparison group for Graph III, provide a percentile conversion, or quantify the uncertainty around an individual difference. The right conclusion is neither that the profile lacks structure nor that its structure answers every score question: Thomas documents a response task and four labeled outputs, while the method needed to interpret finer numerical claims remains unspecified here.

A useful way to keep the documented steps straight is to separate three things: the respondent’s forced choices, the provider’s four named dimensions, and the plotted result. Thomas’s summary identifies the first and names the second, then says the answers create a profile. It does not publish a worked scoring example that would let a reader trace one choice through to a particular plotted point. A reader therefore should not reverse-engineer a score from the adjective task or assume that a visible position corresponds directly to one response.

This distinction helps keep the next reading step precise. The response task and scale names describe how the provider organizes selections into a profile. A report may then place a plot beside prose that gives those dimensions a work-style interpretation. Those are related parts of the document, but a graph’s structure and a narrative’s claims are not the same kind of statement.

Sources: The science behind the Thomas Behaviour assessment

What does the graph show that the narrative adds?

The plotted profile and the prose beneath it do different jobs. In the official “PPA Profile Sample Report 2018,” Graph III organizes results under four dimensions in a displayed profile. The narrative then assigns possible work-style meaning to that pattern, using descriptive language to help a reader consider how it might show up. The report calls itself a guide to be considered with other information. That is a useful boundary: the graph is a structured display, while the paragraph is an interpretation of what such a display may mean.

Neither part records a witnessed work episode. A plotted position is a representation of assessment responses, and a sentence about a possible work tendency is provider-authored interpretation. It may be useful, but it does not add a second measured result merely by appearing next to the graph. The distinction matters because fluent prose can sound more definite than the evidence shown on the page. A reader should be able to tell which part is displayed information and which part is a proposed meaning to examine. (Source: “PPA Profile Sample Report 2018.”)

A practical way to read a narrative sentence is to translate it into a question about a specific episode. For example, imagine a generic report phrase about planning ahead. An invented reflection prompt—not PPA item text or a research case—could be: “In which work situation did I take an action that counts as planning ahead, and what did I actually do?” The question asks for an identifiable action in context, rather than inviting the reader to accept a broad adjective as a fact about every setting.

Then ask what would count against the interpretation. A second invented prompt might be: “Can I recall a comparable situation in which I did not plan ahead, or in which another person supplied the plan?” Such a counterexample does not automatically disprove a recurring tendency; it clarifies its limits and points to conditions that may matter. The useful unit is not a label alone but a claim with a setting, an action, and room for disconfirming experience.

Specificity also makes the narrative more useful. “I plan ahead” can expand into a fixed identity claim; “I usually make a written sequence before a project with several dependencies, but improvise when a request is urgent” is narrower and open to checking. This is an invented illustration, not an assessment result. It shows how to preserve what the report helped the reader notice while avoiding a leap from a general description to an all-purpose conclusion.

The narrative can therefore be a productive starting point: its language may be concrete enough to suggest a reflection the reader had not formulated. Its value lies in helping generate a question, not in settling that question independently. Compare the proposed meaning with episodes, exceptions, and context; keep the wording provisional when the examples are mixed. In that way, graph and prose remain connected, but the reader does not mistake interpretation for additional observation.

Sources: PPA Profile Sample Report 2018; The science behind the Thomas Behaviour assessment

What comparisons does Graph III support?

A plotted level can describe where a person’s dimensions sit in relation to one another within a profile; that is a within-person comparison. A norm-referenced rank instead locates a score relative to scores from a defined group of other people. These are different claims. The opened Thomas API and sample materials show Graph III as a profile display, but they do not identify a Graph III norm group, scale units, or a conversion from a plotted position to a percentile. The visible graph therefore cannot, by itself, be read as a rank against other respondents.

This leaves a third possibility in the debate: comparing whole profiles rather than treating each dimension as a separate score. The 2004 article “Ipsative Assessment Technology and the Personal Profile Assessment by Thomas International” argues that isolated ipsative dimensions are weaker for between-person comparisons, while whole-profile comparisons may be considered in selection. It also makes favorable claims about validity indices for world and Spanish populations. The article page lists Jorge Rodríguez González’s affiliation with Thomas International Spain; that connection is relevant context for weighing the claims, but does not settle whether they are right.

The article is a meaningful counterpoint to a blanket assertion that the PPA has no comparative evidence. Its position is more qualified: separate dimensions and an overall pattern need not support the same kind of comparison, and the author reports favorable validity results. Yet the accessible article page exposes abstract-level claims rather than the information needed to inspect their basis. It does not provide the methods, population details, instrument version, criterion definitions, or numerical indices behind those claims. The distinction between a stated conclusion and inspectable supporting detail is consequential here.

A whole-profile comparison could, in principle, be a different model from ranking one plotted scale. But the article’s general argument does not specify how an individual Graph III point should be interpreted, whether its current display follows the described comparison model, or how a particular person’s point maps onto any reference distribution. Without those particulars, a reader should not convert the sample graph’s height or position into a percentile, nor infer an individual rank from its appearance. That is a limit on what the available materials establish, not proof that no relevant comparative research exists.

The article’s favorable world and Spanish validity statements also cannot simply be transferred to every use of a current Graph III report. Their populations and criteria are not inspectable in the opened page, and the page does not connect a reported index to an individual percentile conversion. The fact that the claims are favorable matters: they challenge categorical dismissal and deserve to be represented accurately. The same evidentiary restraint applies in the other direction, however; an abstract claim without its design and result details cannot document what a specific score means for a particular reader.

These distinctions also affect what a reader may reasonably compare. A profile can invite a descriptive question such as which dimensions appear more prominent for this respondent, subject to the report’s undocumented scoring detail. A norm rank would answer a different question about position among a specified population. A whole-profile model would require its own stated method and comparison evidence. Treating all three as “the graph’s score” would hide the assumptions that distinguish them; a printed vertical location does not supply those assumptions.

The bounded conclusion is that the Graph III materials establish a within-report display, while the available API and sample do not document its norm group, units, or percentile mapping. The 2004 article offers a genuine comparative counterargument, including claims for whole-profile use and favorable validity, with its author’s provider affiliation disclosed. Because the accessible record does not show the underlying methods or indices, it cannot document that a current individual Graph III level is a percentile or rank. A plot’s format and appearance alone do not identify which comparison model generated it.

Sources: PPAAssessment Members, Thomas API Webservices Reference Guide; Ipsative Assessment Technology and the Personal Profile Assessment by Thomas International

What do Thomas's workplace findings actually show?

Thomas’s 2024 technical summary, “The science behind the Thomas Behaviour assessment,” reports several workplace associations that bear on whether the PPA has practical relevance. They are evidence about particular samples and outcomes, not proof that a Graph III narrative predicts one person’s success in a named job. The useful reading is comparative: ask what was measured, in which setting, and whether the direction of the result stays the same across criteria.

In one sample of 132 people, the summary reports correlations between the four profile dimensions and supervisor ratings: Dominance .39, Influence .25, Steadiness .44, and Compliance .14. A correlation describes how two measured variables varied together in that sample; it does not say that a scale caused a rating or determine an individual’s outcome. In a second sample of 79 people, initially matched to a predetermined job description using the provider’s HJA approach, the reported correlations were D .48, I .25, S .44, and C .10. The dimensions did not relate equally to supervisor ratings, and the values differed somewhat across the two groups. The second sample also has a specific design feature: people were initially matched to a predetermined job description, so that result concerns the described matching exercise and its supervisor-rating criterion. It should not be read as a general comparison of all possible jobs. Nor are the two sets of coefficients interchangeable estimates from one pooled sample: their participant counts and reported contexts differ. The summary does not provide enough role composition, selection, rating, or analysis detail to explain those differences or estimate how well an individual prediction would perform. A coefficient alone also leaves practical questions unanswered: it does not give the proportion of successful workers, a threshold for an individual score, or the likely range of ratings for one person.

The next result concerns a different criterion. In a UK furniture-manufacturing company sample of 182 employees, the summary reports an average of 41 lost sales for employees with lower Dominance versus 48 for those with higher Dominance. It also reports conversion of 51.25% for employees with higher Steadiness versus 56.61% for those with lower Steadiness. In that sales setting, lower D and lower S corresponded to better values on the two reported outcomes, though each comparison concerns a different dimension and metric. These figures do not establish why the groups differed, whether their roles and opportunities were comparable, or whether the pattern would recur elsewhere. The two figures also track different parts of a sales process: lost sales and conversion rate. Since the summary gives group averages without enough detail about how the groups were formed or what else differed between employees, readers cannot isolate a scale as the explanation. Even the apparently lower value is only favorable if fewer lost sales or a lower conversion rate is the relevant goal in that role; it says nothing by itself about other responsibilities or overall contribution.

A separate study summary describes 179 people in an organization and a different criterion: manager-rated performance. It reports higher Dominance among top-rated performers and says people with D in their profile had 2.78 times the odds of being in the top-rated group. An odds ratio compares the odds between groups; it is not a 278% increase in performance, a probability for a particular person, or evidence that Dominance produced the rating. This direction contrasts with the lower-D lost-sales finding. The contrast is not a contradiction that can be resolved by declaring one direction universally favorable: selling outcomes in one company and manager ratings in another organization are different measures, jobs, and contexts. “Top-rated” is itself a criterion defined by that organization’s manager ratings, not an independent measure of every aspect of performance. The 2.78 odds ratio compares the reported odds across the profile grouping; without a baseline rate and fuller analysis, it does not tell a reader how likely any individual is to become top-rated. A group association can be informative for a carefully bounded question while remaining too coarse for a personal forecast.

Taken together, Thomas’s reported studies are meaningful counterevidence to the blanket claim that the PPA has no workplace evidence. They also show why a single “better” direction cannot be inferred from a scale label: the association depends on the work setting and the criterion used. The 2024 summary supplies sample sizes and selected coefficients or comparisons, but limited methodological detail, role descriptions, uncertainty estimates, and replication information. These are provider-reported associations at group level; they neither establish causal effects nor show that a particular Graph III narrative sentence accurately describes one respondent or predicts that person’s fit for a job.

Sources: Self- and Observer Reports of Personality

Why might self-description and observed behavior differ?

A self-report records how a person describes themselves; an observer report records how another person describes them. If Graph III’s self-description and a colleague’s account differ, the difference is worth examining, but it does not show that the two perspectives must be incompatible—or that either one is automatically correct.

The 2025 Annual Review synthesis “Self- and Observer Reports of Personality” reports comparable average scores and fairly high convergence between self and observer reports for full-length Big Five and HEXACO questionnaires. Convergence varied by trait and was somewhat lower for cooperation-related traits. The review concerns those established models and questionnaire measures; it is not a study of Thomas PPA Graph III, workplace colleagues, or the accuracy of a particular report. It therefore counters the broad assumption that self and observer accounts inevitably diverge, while providing no direct validation of the PPA narrative.

Consider a hypothetical exchange, offered only as an illustration. A report phrase might describe someone as inclined to make decisions independently; a colleague recalls a project where the person repeatedly asked the team to choose the next step. Before treating these as mutually exclusive verdicts, the two could specify the episode: What decision was at stake, who held the relevant information, and what deadline or responsibility shaped the choice? Perhaps the person worked independently on familiar tasks but sought input when consequences affected the group. That possibility is a question to check against experience, not an explanation established by the report or the colleague’s recollection. A different task could produce another account: perhaps the same person proposes a course of action alone when the problem is familiar, but invites joint choice when responsibilities overlap. This remains a hypothetical possibility; neither the review nor the invented exchange says that this is what happened for any real person.

A fair interpretation preserves both accounts long enough to make them precise. Ask what each person observed, under which task demands, and whether there are comparable episodes in the other direction. A colleague may see only a slice of someone’s work, while a self-description may draw on situations the colleague never witnessed; that is a possibility to check, not presume. Then decide whether the difference reflects context, a difference in what each person noticed, or a real disagreement that needs more evidence. Agreement between self and observer reports in Big Five and HEXACO research makes corroboration plausible in some settings; it cannot certify that this PPA description fits, or settle which account should guide a particular work decision.

Sources: Self- and Observer Reports of Personality

What kind of evidence would justify a stronger reading?

A stronger reading is justified only when evidence matches the exact claim and the decision someone wants to make from it. The joint AERA, APA, and NCME Standards for Educational and Psychological Testing define validity around evidence and theory supporting a particular interpretation of scores for a particular use. That definition does not label a whole test simply “valid” or certify every sentence in a report. Applied here, evidence that a score can support a bounded reflection would not automatically support using it to rank people, select an employee, or forecast job performance. The stronger the consequence, the closer the evidence must fit the proposed interpretation and use.

For a claim about what a Graph III score means, documentation would need to connect the score’s construction to the proposed interpretation, including the relevant version and the population in which that interpretation was studied. A reader might then consider whether “higher” means more of a defined construct, a relative pattern among this person’s scales, or something else. Consistency is a separate matter: reliability describes how dependable scores are under specified conditions, while precision concerns how much uncertainty surrounds a score. The Standards call for appropriate reliability and precision evidence for each intended use. If two plotted values sit close together, a fine distinction needs enough precision to show that the apparent gap is meaningfully distinguishable from measurement error. A stable score cannot rescue an interpretation whose meaning has not been supported.

A percentile or rank needs a stated comparison basis: a defined norm group, relevant administration conditions, and a documented method connecting the score to that group’s distribution. Without those details, a graph’s position is not itself evidence of where someone stands among other people. Likewise, an employment inference needs evidence that fits the role, the population being considered, the work outcome used as a criterion, and the purpose of the decision. Findings about one outcome or group cannot simply be transferred to a different role or to a high-stakes decision. Fairness and appropriateness also matter: even evidence that supports an interpretation does not by itself show that using it for a particular decision is fair or suitable.

Thomas’s “PPA Profile Sample Report 2018” advises treating the report as a guide alongside other information. That is provider direction, not empirical proof that any particular combination of information improves accuracy. It does point toward a useful boundary: keep a report interpretation open to relevant context rather than letting a graph settle a consequential judgment by itself. This framework does not establish that Graph III is invalid; relevant documentation could support a stronger, carefully bounded claim. Before relying on a fine distinction or work conclusion, ask: what evidence supports this exact interpretation for this specific decision, and how much uncertainty remains? Documentation should also make clear whether findings apply to the current version and how uncertainty was estimated, rather than relying on a familiar instrument name alone. A general personality report also does not establish a clinical diagnosis; that is outside this report-reading question.

Sources: Standards for Educational and Psychological Testing; PPA Profile Sample Report 2018

What should you do with Graph III next?

Choose one phrase from the narrative and one work question it might help you examine, such as when a collaboration friction appears. Compare it with one episode and one exception: what happened, what did you do, and what differed when the pattern did not fit? Treat this as reflection, not a test of graph accuracy. Keep attention on conditions and collaboration, not labels or ranking. If examples are mixed, keep the description conditional and discuss the episode with a coach or colleague who can help you consider the work context. For a separate prompt about how you decide, plan, collaborate, handle conflict, adapt, or learn, the [Work Pattern Report](/assessment) offers a low-stakes ten-continuum reflection. It is not validated for employment decisions, a Thomas PPA equivalent, or a job recommendation. Use it to name reflection questions, not to confirm Graph III. Browse [personality report topics](/topics) for more. In your next conversation, ask when the pattern appeared, what shaped it, and what would make collaboration easier; share a concrete example and invite the other person’s observation so you can discuss one practical adjustment.

Questions readers ask