Attitude Polling as Research Foundation

The Scout Before the Army

You don't move an army without scouts. The terrain might look clear from a distance and be impassable up close. The village that appears friendly might be garrisoned by the enemy. Without ground-level intelligence, the best-designed campaign fails at contact.

For Bernays, attitude polling is the scout work that precedes any consent-engineering campaign. You cannot design an effective persuasion strategy for an audience whose existing attitudes you do not know. More precisely: you cannot know which of your planned interventions will work with the current of existing opinion and which will run against it. The existing attitude landscape is the terrain. Polls map the terrain before the campaign moves.¹

This is a significant claim about what polls are for. The popular image of polling is democratic and descriptive — polls tell you what people think, providing a mechanism for public preferences to be known. Bernays uses polls in an instrumental and prescriptive role — polls tell you where people are now, so you can design the campaign that moves them from there to where you want them. The measurement serves the movement.

The Research Phase in Engineering of Consent

Bernays' eight-step engineering of consent methodology places research at Step 2, immediately after defining the goals. This sequencing is not incidental. Before any strategy, theme, or tactic is designed, the practitioner must know:

What do the relevant publics currently believe about the subject?
How strongly are those beliefs held? Are they peripheral (malleable) or identity-attached (logic-proof)?
Which groups are already favorable? Which are hostile? Which are unformed?
Who are the relevant opinion leaders for each group?
What stereotypes and clichés are currently active around the issue?
What events or developments have recently shaped the opinion landscape?²

Without this knowledge, campaign design is guesswork. With it, the practitioner can identify which existing attitudes to work with (amplifying favorable dispositions), which to route around (avoiding direct confrontation with logic-proof compartments), and which to reframe (changing the cliché without attacking the belief).

The research finding also determines which group leaders to target first — not the group leaders of hostile constituencies (who will be the last to shift) but the group leaders of unformed or weakly favorable constituencies (where movement is possible and cascade effects can follow).

Polls as Terrain Map vs. Polls as Force

Here is where the concept bifurcates, and Bernays does not fully acknowledge the split.

Polls as terrain map: The legitimate research function. A poll conducted before campaign design reveals the existing attitude landscape and informs strategy. The poll affects the campaign but not the public — it is a measurement tool that feeds into private decision-making.

Polls as force: The less-acknowledged application. Once a poll is published, it enters the river of interacting forces as a social signal. A poll showing 70% support for a position is not just data — it is news, and as news it functions as evidence of what group members believe, which triggers the herd instinct response: if most people in my relevant groups hold this position, perhaps I should too. The published poll is a manufactured endorsement — real in the sense that the polling happened, but shaped in its effects by how it is framed, which questions were asked, which results were highlighted, and which were not.

Bernays acknowledges in his essay on attitude polls that genuine democratic leaders must lead ahead of public opinion rather than simply following polls — that the poll-driven politician who adjusts positions based on what 51% of respondents say is abdicating leadership responsibility for a kind of inverted manipulation, using the public to avoid governing. This is a genuine insight about democratic leadership. But it sits uneasily with the rest of his framework, which deploys polling as an instrument for engineering rather than following opinion.³

The Structural Paradox

The deeper paradox in attitude polling as research foundation is that it treats public opinion as the input to a process whose output is also public opinion. The sequence:

Poll existing attitudes (research)
Design campaign to move those attitudes in a desired direction (strategy)
Deploy manufactured events, group leader endorsements, cliché replacement (tactics)
Attitudes shift
New poll measures the shifted attitudes (evaluation)

The practitioner uses step 1 to make step 4 more likely — they are not discovering what people think and accommodating it; they are discovering where the leverage points are and applying force there. The final step (step 5) measures the success of the engineering, not the authentic preferences of an undisturbed public.

This process is invisible to the people whose attitudes are being polled and subsequently moved. From inside the experience, their opinions formed naturally, their preferences are their own, and the poll simply documented what they thought. The manufacturing happened between the first poll and the last — in the design phase where the existing attitude landscape was used as a map to plan the route to the desired destination.

Evidence

The "Attitude Polls" essay: Bernays' own essay on polling distinguishes between polls as democratic instruments (measuring existing preference to guide representative government) and polls as research tools for consent engineering (measuring existing attitude to design targeted persuasion). He is more candid about this dual use than critics usually credit him with — he acknowledges both functions without fully confronting the tension between them.⁴

The velvet campaign research phase: Before designing the velvet campaign, Bernays conducted research on American fashion channels — which publications set fashion trends, which opinion leaders (Paris couturiers, American fashion editors, department store buyers) were the key nodes in the distribution chain, which months produced the highest impact for fashion-forward material, and what the existing attitude toward velvet was. This research determined the sequence: Paris couturiers first, fashion editors second, department stores third, consumers last — because that was how the fashion influence chain actually ran. The poll equivalent here is not a formal opinion survey but a structured mapping of the terrain before the campaign moved.⁵

The research-before-strategy requirement across campaigns: Every Bernays case study in the engineering of consent methodology includes a research phase before strategy design. The NAACP Atlanta conference was preceded by research identifying Southern Baptist ministers as the key group leaders for Southern white Christian opinion on racial matters. Light's Golden Jubilee was preceded by research identifying Edison as the optimal symbolic figure for electricity's public image — not just famous, but already carrying the specific associations (inventor, American genius, self-made man) the campaign needed. The overt act must be designed to work with existing associations, and existing associations must be mapped before the design.⁶

Tensions

The push poll problem: If polls can function as social forces rather than just measurement tools, then the methodology for designing "research polls" and "influence polls" converges. A poll that asks leading questions (Was Edison the greatest American inventor?) and then releases results ("70% of Americans believe Edison is the greatest inventor!") is simultaneously measuring and manufacturing. Bernays does not discuss this boundary, but his framework provides no principled stopping point before the push poll — it is just an overt act using the poll format as the manufactured event.

The democratic leadership tension: Bernays argues that genuine democratic leadership requires leaders who sometimes act ahead of public opinion rather than following polls — that the poll-driven leader abdicates moral responsibility for a form of inverterd pandering. This is a genuine principle. But it sits next to a professional methodology in which the PR counsel deploys polling specifically to understand where public opinion currently is before designing the campaign that will move it. One cannot simultaneously hold that leaders should lead ahead of polls and that practitioners should use polls to design campaigns targeting the existing opinion landscape. The tension is real and unresolved in Bernays' work.⁷

The research-as-weapon problem: Research conducted before a persuasion campaign is technically neutral — it maps what people believe. But the insights from the map are applied asymmetrically: the practitioner's client gets the map, the public does not know a map was made or a campaign designed. The public's attitudes, revealed through polling, become the strategic resource of the party with the resources to commission the poll and design the campaign. Research democratizes nothing if only one side of an interaction has access to it.

Cross-Domain Handshakes

The two-line version: attitude polling as research foundation is intelligence-gathering applied to social terrain — the same discipline that appears in military intelligence and investigative practice, now applied to the pre-campaign mapping of public opinion.

History: Maratha Intelligence and Spy Network — Bahirji Naik's intelligence apparatus (knowing the terrain, fort layouts, army dispositions before operations began) and Bernays' attitude research phase are structurally identical: you map the terrain before moving, because misreading the terrain is costlier than the map. The parallel runs all the way through: Naik used human networks to map physical and social terrain; Bernays used polls and group-structure analysis to map opinion terrain. The insight the two produce together: research-before-action is not a methodology specific to military or social domains — it is a general principle wherever the environment is adversarial and misreading it is expensive.
Psychology: Public Opinion as Interaction — The interacting forces model says public opinion is a product of organized forces; attitude polling maps the current output of those forces at a given moment. Polling alone does not reveal the forces producing the opinion — it reveals the downstream product. The limitation: if you only poll, you know where people are but not what produced that position and therefore not where the leverage points for change are. Bernays' research methodology combines polling (measuring outputs) with group-structure analysis (mapping the forces producing those outputs) — the combination produces actionable intelligence, not just description.
Cross-domain: Engineering of Consent — Attitude polling is Step 2 of the eight-step engineering of consent methodology. Without the research phase, Steps 3-8 (strategy, theme, organization, timing, tactics, budget) cannot be designed with any precision — they would be guessing. The polling step is what makes the engineering systematic rather than intuitive. The handshake: every claim Bernays makes about the engineering being a planned, scientific methodology rests on the research phase being genuinely diagnostic. If the research only confirms what the practitioner already believed, the methodology is rationalized intuition rather than applied social science.

The Live Edge

The Sharpest Implication

Attitude polling was designed as a democratic instrument — a mechanism for governments to know what citizens prefer, so representative institutions could represent those preferences. Bernays converts it into a pre-campaign intelligence tool: the poll maps the terrain so the engineering of consent can be applied more precisely. This conversion does not require any corruption of the polling methodology. The same questions, the same sampling, the same analysis can be used for both purposes. The difference is entirely in what happens after the results come back: the democratic use informs policy to match preferences; the engineering use informs campaigns to change preferences. The public that answered the questions cannot know which use their answers will serve. And in the post-poll world where their answers have been used to design a campaign that moves their opinions, the follow-up poll will show their opinions have "moved" — and nobody inside that process can distinguish organic preference change from engineered attitude shift. The measurement and the manipulation use the same instrument. The democratic function and the engineering function are identical at the level of practice and invisible to the people measured.

Generative Questions

If polls can function simultaneously as measurement tools and as social forces (published polls produce herding effects), is there a polling methodology that minimizes the social-force effect while preserving the measurement function? Would polling need to be conducted and kept private to avoid contaminating the opinion environment it is trying to measure?
Bernays argues democratic leaders should lead ahead of polls rather than following them. But he also argues the PR counsel should use polls to identify the current opinion landscape before designing campaigns. Is there a coherent position that resolves this tension — a principle for when polls should lead decisions and when decisions should lead polls?
The push-poll is the boundary case where polling converts entirely into manufactured event. What is the precise mechanism by which a measurement tool becomes an influence tool? Is the push poll a category error, or is it the honest completion of a trajectory already present in legitimate polling methodology?

Connected Concepts

Engineering of Consent — the methodology that deploys polling as its research foundation
Public Opinion as Interaction — the theory of opinion formation that determines what polls are actually measuring
PR Counsel as Profession — the professional who commissions and interprets the research
Manufactured Event / Overt Act Theory — the instrument the research phase is designed to deploy most effectively

Open Questions

Does contemporary research on polling and opinion formation (including effects of releasing poll results) confirm or challenge Bernays' use-case distinction between measuring and manufacturing opinion?
Is there a meaningful distinction between attitude polling commissioned by a party with a persuasion interest and attitude polling commissioned by a disinterested researcher? What institutional forms would preserve that distinction?